there were a lot of events recorded by gharchive.org of which 2,149,484 were push events containing 3,286,551 commit messages that amount to 247,419,739 characters filtered with words.py@e23d022007... to these 61 messages:
Hologram Touchup (Init savings edition) (#74793)
Hologram generation currently involves a bunch of icon operations, which are slow. Not to mention a series of get flats for the human models, which is even worse.
We lose 0.05 seconds of init to em off just the 2 RCD holograms. it hurts man.
So instead, let's use filters and render steps to achive the same effect.
While I'm here I'll dim the holo light and make it blue, make the hologram and its beam emissive (so they glow), and do some fenangling with move_hologram() (it doesn't clear the hologram off failure anymore, instead relying on callers to do that) to ensure holocalls can't be accidentially ended by moving out of the area.
Ah and I added RESET_ALPHA to the emissive appearance flags, cause the alpha does override and fuck with color rendering, which ends up looking dumb. If we're gonna support this stuff it should be first class not accidential.
While I'm here (since holograms see static) lets ensure the static plane is always visible if you're seeing through an ai eye.
The old solution was limited to applying it to JUST ais, which isn't satisfactory for this sort of thing and missed a LOT of cases (I didn't really get how ai eyes worked before I'ma be honest)
I'm adding a signal off the hud for it detecting a change in its eye here. This is semi redundant, but avoids unneeded dupe work, so I'm ok with it.
The pipeline here is less sane then I'd like, but it works and that's enough
More pretty, better ux, static works
🆑 add: Holograms glow now, pokes at the lighting for holocalls in general a bit to make em nicer. qol: You can no longer accidentally end a holocall (as a non ai) by leaving the area. Felt like garbage fix: Fixes static rendering improperly if viewed by a non ai /🆑
Refactors and defuckulates dbcore. Adds support for min_threads rustg setting, Reduce query delay, Make unit tests faster (#74852)
dbcore was very fuckulated.
It had 3 lists of queries, but they all had their own current_run style
list to support mc_tick_check (as it was already being done before with
the undeleted query check, so i can understand why they cargo culted
mirrored the behavior) This was silly and confusing and unneeded given
two of those loops can only process at most 25 items at a time on
default config, plus these were cheap operations (ask rustg to start
thread, ask rustg to check on thread).
Because of the confusingness of the 6 lists for 3 query states, The code to run pending/queued queries immediately during world shutdown was instead looking at the current_run list for active queries, meaning those queries got ran twice.
The queued query system only checked the current active query count in fire(), meaning even when there was nothing going on in this subsystem new queries had to wait for the next fire() to run (10 ticks, so 500ms on default config)
Those have all been fixed.
the config BSQL_THREAD_LIMIT has been renamed to
POOLING_MAX_SQL_CONNECTIONS and its default was lowered to match
MAX_CONCURRENT_QUERIES .
added a new config POOLING_MIN_SQL_CONNECTIONS, allowing you to
pre-allocate a reserve of sql threads.
The queue processing part of SSdbcore's fire() has been made to not obey mc_tick_check for clarity and to make the following change easier to do:
If there is less than MAX_CONCURRENT_QUERIES in the active queue, new
queries activate immediately.
(its ok that there are two configs that kinda do the same thing, POOLING_MAX_SQL_CONNECTIONS maps to max-threads in the mysql crate, and it seems to only be a suggestion, meanwhile MAX_CONCURRENT_QUERIES can't do anything during init, which is when the highest amount of concurrent queries tend to happen.)
🆑 config: database configs have been updated for better control over the connection pool server: BSQL_THREAD_LIMIT has been renamed to POOLING_MAX_SQL_CONNECTIONS, old configs will whine but still work. fix: fixed rare race condition that could lead to a sql query being ran twice during world shutdown. /🆑
I have not tested this pr.
Adds admin alert for revs created through traitor panel (#74862)
So like, using traitor panel to make revs doesn't work.
Revolutions live and die, currently, by the revolution ruleset datum dynamic creates. It manages the hostile environment and also processes to check whether either side should be winning or not.
This means that the revolutionary buttons in the traitor panel are kind of noob-admin-bait. You press it for a funny revolution and then you realize it's screwed when all the heads are dead and everyone's stumbling around cluelessly
This has a proper solution, albeit somewhat difficult - separate out the revolution from the ruleset, make admin spawned revs create a revolution. I can do this but it's a lot of effort and this works in the meanwhile
Pops up a TGUI alert when an admin presses "add revolutionary" in traitor panel when there is no ongoing revolution. Simply enough, gives them an alert that it will not work correctly. Lets them decide whether they want to deal with that. (Because you can manually deal with it via proc calls, if you've got code smarts.)
Stops admins from stumbling into the same trap without warning.
Can be removed in the future easily when revs are coded better.
🆑 Melbert admin: Adds a warning that spawning revs via traitor panel will not function as expected. /🆑
Makes a whole bunch of wooden objects flammable (#74827)
This whole PR started because I realized that baseball bats are not actually flammable which I found weird, then I looked at a whole bunch of other stuff that really should be flammable but also isn't.
Makes wooden objects behave slightly more consistently? Honestly, most of these seem like oversights to me.
🆑 balance: The following structures are now flammable: Picture frame, fermenting barrel, drying rack, sandals, painting frames, paintings, spirit board, notice board, dresser, displaycase chassis, wooden barricade balance: The following items are now flammable: Baseball bat, rolling pin, mortar, coffee condiments display, sandals, wooden hatchet, gohei, popsicle stick, rifle stock /🆑
You can't instantly resist out of an unlocked labor camp teleporter if you are handcuffed (#73983)
If you are restrained, and placed into an unlocked labor camp teleporter, you cannot instantly resist out of it. However the resist timer is cut in half while unlocked.
Getting someone into the gulag teleporter is an incredibly un-necessary pain in the rear because simply spamming resist turns it into a game where you have to shove them in, then really quick go over to the computer and slam the lock button. This is... kinda lame. A lot of new player security officers get got by this, and I think it's sad. Inb4 "Skill issue"
🆑 Melbert balance: If you are handcuffed, you can't instantly resist out of an unlocked labor camp teleporter (however, resist time is halved). /🆑
[Needs TM] The Accelerataning (#1781)
Gone are the days of spam clicking buttons to move faster in a direction, with this PR, ships now accelerate constantly (as long as you have fuel and don't touch the throttle) in a direction you set, leading to a much smoother flight experience. I imagine it's going to be a bit tougher to thread gaps, but flying a spaceship is quite literally rocket science. So.
Also actually makes the minimum and maximum speed work, and adjusts them to a more tolerable level.
Eliminates the ability to cheese high speeds by spamming the accelerate button, and also makes the flight experience much more pleasant as you don't have to spam click to move a decent speed.
🆑 add: A new system for ship flight, where you only point a direction and set the throttle to change your speed, reducing the need for spam-clicking. fix: There's now a maximum and minimum speed, 600spm and 0.01spm, respectively. The limits have been broken all this time. /🆑
Signed-off-by: Mark Suckerberg [email protected]
[MIRROR] Fixes a runtime in simple_animal/hostile [MDB IGNORE] (#20588)
- Fixes a runtime in simple_animal/hostile (#74706)
Attempting to fix this flaky test that has been cropping up from the Icebox tests. It is annoying.
From what I can tell, the mob was getting qdeleted while it was doing its loop of finding a target. This can happen at any time, because many simple mobs (including the one causing the issues) get qdeleted on death.
Added some more checks to make sure we don't do certain actions if the mob gets qdeleted midway through execution of its AI routine. It really could happen anywhere so we must be vigilant.
create_and_destroy: [02:24:31] Runtime in stack_trace.dm,4: addtimer called with a callback assigned to a qdeleted object. In the future such timers will not be supported and may refuse to run or run with a 0 wait (code/controllers/subsystem/timer.dm:583)
proc name: stack trace (/proc/_stack_trace)
src: null
call stack:
stack trace("addtimer called with a callbac...", "code/controllers/subsystem/tim...", 583)
addtimer(/datum/callback (/datum/callback), 300, 8, null, "code/modules/mob/living/simple...", 595)
the demonic watcher (/mob/living/simple_animal/hostile/asteroid/ice_demon): GainPatience()
the demonic watcher (/mob/living/simple_animal/hostile/asteroid/ice_demon): GiveTarget(the mi-go (/mob/living/simple_animal/hostile/netherworld/migo))
the demonic watcher (/mob/living/simple_animal/hostile/asteroid/ice_demon): FindTarget(/list (/list))
the demonic watcher (/mob/living/simple_animal/hostile/asteroid/ice_demon): AIShouldSleep(/list (/list))
the demonic watcher (/mob/living/simple_animal/hostile/asteroid/ice_demon): handle automated action() at stack_trace.dm:4
On top of that, there is signal handling in place to LoseTarget() when a
mob that is already a target gets qdel'd and sends
COMSIG_PARENT_QDELETING. Shown below.
However there is nothing stopping a target that is not null but that has been qdeleted from being considered as a target in the first place.
This PR just aims to fix that problem by making sure that a) a hostile ai that gets qdeleted midway through does not keep doing stuff that can cause issues and b) an atom that is being qdeleted never makes its way into the targets list of a hostile ai.
Simple mobs/AI are due for a wider refactor honestly but this really ought to be done in the meantime so we don't get spammed by CI failures over nonsense.
Fixes tgstation/tgstation#73032 Fixes tgstation/tgstation#74266 Fixes Skyrat-SS13/Skyrat-tg#18964 Fixes Skyrat-SS13/Skyrat-tg#19749 Fixes Skyrat-SS13/Skyrat-tg#18964 Fixes Skyrat-SS13/Skyrat-tg#19322 Fixes Skyrat-SS13/Skyrat-tg#18974 Fixes Skyrat-SS13/Skyrat-tg#19296 Fixes Skyrat-SS13/Skyrat-tg#19294
Bugfix, stops the icebox test from failing as much.
🆑 fix: fixes hostile mobs sometimes being able to target an atom that has been marked for deletion and then becoming confused, and in a similar vein fixes mobs sometimes still running their AI while being marked for deletion. /🆑
Co-authored-by: san7890 <the@ san7890.com>
- Fixes a runtime in simple_animal/hostile
Co-authored-by: Bloop [email protected] Co-authored-by: san7890 <the@ san7890.com>
[MIRROR] Icemoon Hermit Ruin Active Turf Fix - For Real This Time [MDB IGNORE] (#20325)
- Icemoon Hermit Ruin Active Turf Fix - For Real This Time (#74476)
In #74306, I thought I knew what the cause was, and I both attempted a potential fix and made tracking it easier. The fruits of my labor paid off, I know exactly what caused it now.
Basically, the demonic portal will scrape away all turfs in a 5-tile
radius on its Initialize(), and if a spawner spawned right next to the
hermit ruin... it would count it as a mineral turf and scrape it away as
well. That's so fucking silly. At least we know now.
The fix is to just make those tiles unscrapeable, which is accomplished
via another turf_flag and filtering those out in the Initialize() of
the demonic portals.
I also cleaned up the calls to scrapeaway being null, which is really
weird because it just defaulted to the normal proc behavior. Naming the
arguments instead does the same thing (I checked)
- Icemoon Hermit Ruin Active Turf Fix - For Real This Time
Co-authored-by: san7890 [email protected]
[MIRROR] IceBoxStation More Active Turf Fixes [MDB IGNORE] (#20339)
- IceBoxStation More Active Turf Fixes (#74474)
This didn't show up in my testing for #74410. I hate it here.
I am a monkey trapped next to a computer playing whackamole with this fucking chasms and active turfs. one day i will be free.
nothing that should concern players
- IceBoxStation More Active Turf Fixes
Co-authored-by: san7890 [email protected]
[ADD] website_cf_turnstile: add cloudflare turnstile support
This module allows to add secret key to add the turnstile captcha on each snippet website_form.
A friendly, free CAPTCHA replacement Turnstile delivers frustration-free, CAPTCHA-free web experiences to website visitors. Turnstile stops abuse and confirms visitors are real without the data privacy concerns or awful UX that CAPTCHAs thrust on users.
closes odoo/odoo#116252
Signed-off-by: Jérémy Kersten [email protected]
[MIRROR] Mafia rebalance and backend refactor [MDB IGNORE] (#20631)
- Mafia rebalance and backend refactor (#74640)
Turns all Mafia abilities into datums, instead of being a bunch of shitcode on every single job. This means it's easier to add new roles Gives new names to some defines (such as the signal order, to make it easier to tell when something is fired) Adds support for modular Mafia jobs with their abilities being in a certain order (Escort is now properly first). De-snowflakes Changeling killing abilities and day voting, they're now actions that are tallied when necessary.
Turns time vars into defines Generalizes a lot of behavior for abilities, now all abilities can properly undo their action at night
Fixes problems with the UI (Thoughtfeeder had 2 buttons during night and they overlapped with names, that's been fixed).
- Doctor/Officer can now protect themselves 1 night, because it gives them a way to protect themselves.
- Lawyer/Warden/Ect now choose their abilities at night, rather than the day before. The suspense building up towards the end of the night is part of the game, telling you that it happened at the very start is quite lame (in the case of Lawyer, anyway).
- Admin setup now uses TGUI instead of html inputs.
- Cut night time by like, 5 seconds, because I found it a little long lol.
- HoP doesn't count as votes to win until they reveal, because it makes no sense an unrevealed HoP has their unrevealed votes tallied. I also like those 1v1 Mayor V. Evil scenarios where dead chat goes crazy, and hope to replicate that here.
- Mafia now needs 6 people to start instead of 4, because 4 players is just not enough to play a Mafia round that will do anything but annoy people.
- The game no longer ends if it's in a standoff with 1 Town, 1 Mafia, and 1 Neutral, as you've got a kingmaker and they should decide who wins.
Every time night starts/ends, it checks the entire GLOB.airlocks for
doors with the "mafia" ID. This is stupid.
Rework check_victory() to make it make more sense, and be more fun
for players.
A visible death animation?
I want to use something similar to admin popup for messages about people
being on stand, and decluttering the UI in general
Also more use of balloon alerts instead of to chat messages for
everything.
Also also, making the UI more responsive to players. Button should be
red when a player is selected, so they know that's who they've selected,
if they want to unselect.
Are votes public when you first cast them? They shouldn't be wtf.
Can we also make the description for roles not be a to chat message? It
can just say when you hover over the '?' come on.
User-written wills instead of auto-generated, and able to send them in
chat
Add support for roleblock-immune roles
Updates a lot of old code to modern standards Makes it considerably easier to work with Mafia and add new roles Makes things less prone to breaking as easily. Code also looks a lot cleaner now.
🆑 refactor: [Mafia] All Mafia abilities have been overhauled in the backend, it's now much easier to understand what each role's ability can do and how it works. admin: [Mafia] Admin setup of Mafia is now in TGUI balance: [Mafia] Doctors/Officers can protect themselves once per game. Be careful around them! fix: [Mafia] Thoughtfeeder's UI buttons at night won't overlap with eachother. fix: [Mafia] HoP's votes now actually matter, instead of being purely visual. qol: [Mafia] Lawyers, Wardens, etc. now perform their night ability at night, instead of the day prior. qol: [Mafia] Night time now lasts 40 seconds instead of 45. /🆑
- Mafia rebalance and backend refactor
Co-authored-by: John Willard [email protected]
windows: ignore empty PATH elements
When looking up an executable via the _which function, Git GUI
imitates the execlp() strategy where the environment variable PATH
is interpreted as a list of paths in which to search.
For historical reasons, stemming from the olden times when it was uncommon to download a lot of files from the internet into the current directory, empty elements in this list are treated as if the current directory had been specified.
Nowadays, of course, this treatment is highly dangerous as the current
directory often contains files that have just been downloaded and not
yet been inspected by the user. Unix/Linux users are essentially
expected to be very, very careful to simply not add empty PATH
elements, i.e. not to make use of that feature.
On Windows, however, it is quite common for PATH to contain empty
elements by mistake, e.g. as an unintended left-over entry when an
application was installed from the Windows Store and then uninstalled
manually.
While it would probably make most sense to safe-guard not only Windows
users, it seems to be common practice to ignore these empty PATH
elements only on Windows, but not on other platforms.
Sadly, this practice is followed inconsistently between different software projects, where projects with few, if any, Windows-based contributors tend to be less consistent or even "blissful" about it. Here is a non-exhaustive list:
Cygwin:
It specifically "eats" empty paths when converting path lists to
POSIX: https://github.com/cygwin/cygwin/commit/753702223c7d
I.e. it follows the common practice.
PowerShell:
It specifically ignores empty paths when searching the `PATH`.
The reason for this is apparently so self-evident that it is not
even mentioned here:
https://learn.microsoft.com/en-us/powershell/module/microsoft.powershell.core/about/about_environment_variables#path-information
I.e. it follows the common practice.
CMD:
Oh my, CMD. Let's just forget about it, nobody in their right
(security) mind takes CMD as inspiration. It is so unsafe by
default that we even planned on dropping `Git CMD` from Git for
Windows altogether, and only walked back on that plan when we
found a super ugly hack, just to keep Git's users secure by
default:
https://github.com/git-for-windows/MINGW-packages/commit/82172388bb51
So CMD chooses to hide behind the battle cry "Works as
Designed!" that all too often leaves users vulnerable. CMD is
probably the most prominent project whose lead you want to avoid
following in matters of security.
Win32 API (CreateProcess())
Just like CMD, `CreateProcess()` adheres to the original design
of the path lookup in the name of backward compatibility (see
https://learn.microsoft.com/en-us/windows/win32/api/processthreadsapi/nf-processthreadsapi-createprocessw
for details):
If the file name does not contain a directory path, the
system searches for the executable file in the following
sequence:
1. The directory from which the application loaded.
2. The current directory for the parent process.
[...]
I.e. the Win32 API itself chooses backwards compatibility over
users' safety.
Git LFS:
There have been not one, not two, but three security advisories
about Git LFS executing executables from the current directory by
mistake. As part of one of them, a change was introduced to stop
treating empty `PATH` elements as equivalent to `.`:
https://github.com/git-lfs/git-lfs/commit/7cd7bb0a1f0d
I.e. it follows the common practice.
Go:
Go does not follow the common practice, and you can think about
that what you want:
https://github.com/golang/go/blob/go1.19.3/src/os/exec/lp_windows.go#L114-L135
https://github.com/golang/go/blob/go1.19.3/src/path/filepath/path_windows.go#L108-L137
Git Credential Manager:
It tries to imitate Git LFS, but unfortunately misses the empty
`PATH` element handling. As of time of writing, this is in the
process of being fixed:
https://github.com/GitCredentialManager/git-credential-manager/pull/968
So now that we have established that it is a common practice to ignore
empty PATH elements on Windows, let's assess this commit's change
using Schneier's Five-Step Process
(https://www.schneier.com/crypto-gram/archives/2002/0415.html#1):
Step 1: What problem does it solve?
It prevents an entire class of Remote Code Execution exploits via
Git GUI's `Clone` functionality.
Step 2: How well does it solve that problem?
Very well. It prevents the attack vector of luring an unsuspecting
victim into cloning an executable into the worktree root directory
that Git GUI immediately executes.
Step 3: What other security problems does it cause?
Maybe non-security problems: If a project (ab-)uses the unsafe
`PATH` lookup. That would not only be unsafe, though, but
fragile in the first place because it would break when running
in a subdirectory. Therefore I would consider this a scenario
not worth keeping working.
Step 4: What are the costs of this measure?
Almost nil, except for the time writing up this commit message
;-)
Step 5: Given the answers to steps two through four, is the security measure worth the costs?
Yes. Keeping Git's users Secure By Default is worth it. It's a
tiny price to pay compared to the damages even a single
successful exploit can cost.
So let's follow that common practice in Git GUI, too.
Signed-off-by: Johannes Schindelin [email protected]
add alpaca gpt4 dataset (#2610)
The inputs can be quite a lot of different versions of no input,
therefore don't use the input column for that.
In some cases the text in input is already in the instruction, in
these cases, we also don't use the input column.
I am not quite sure how to concatenate the instruction and the input
column. In most cases it seems fine to just replace last appearance of
., ! or ? with a colon, e.g.:
Instruction: Identify the odd one out.
Input: Twitter, Instagram, Telegram
or
Instruction: How dense is a given material?
Input: Steel
But we also have some questions like:
Instruction: Given the following synopsis, what is the moral lesson of this story?
Input: Once upon a time, there was a poor young boy who wanted some candy. He begged his father for money to buy it, but his father said no and ordered him to go to bed. As he was going to bed, the boy saw a five-dollar bill on the counter, which he took and bought the candy.
Where this might not be the best case. Either way, I think the this one token will not make significant difference the model and therefore I just concatenate instruction and input with a space.
[MIRROR] Polishes some side sources of light and color [MDB IGNORE] (#19860)
- Polishes some side sources of light and color (#73936)
Circuit floors glow! but it looks like crap cause it's dim and the colors are washed out. I'd like to make them look nicer. Let's make them more intense and longer range, and change the colors over to more vivid replacements.
While I'm here, these should really use power and turn on and off based off that. Simple enough to do, just need to hook into a signal (and add a setter for turf area, which cleans up other code too).
Desklamps look bad. They're fullwhite, have a way too large range.Crummy. Let's lower their lightrange from 5 to 3.5, and make the ornate ones warmer, and the more utilitarian ones cooler. The clown one can be yellow because it's funny
I'm renaming a color define here so I'm touching more files then you'd expect
Increases the light range of request consoles, status displays, newscasters, and air alarms (keycard machines too, when they're awaiting input at least) Increases the brightness of air alarms, I think they should be on par with apcs, should be able to tell when they're good/bad. Increases the brightness of vending machines (I want them to light up the tiles around them very lightly, I think it's a vibe)
Fixes a bug with ai status displays where they'd display an emissive even if they didn't have anything on their screen, looking stupid. This was decently easy but required a define. Looked really bad tho
Pretty
Circuit Floors
Old
New
Desk Lights
Old
New
The niknack changes are more minor so I'm not gonna grab photos for them. I can if you'd like but I don't think it's necessary. Mostly a vibes in dark spaces sorta thing
🆑 add: I made circuit floors brighter and more vivid. add: Made air alarms, vending machines, newscasters, request consoles, status displays and keycard machines slightly "brighter" (larger light range, tho I did make air alarms a bit brighter too) add: Tweaked desklamps. Lower range, and each type gets its own coloring instead of just fullwhite. fix: AI displays are no longer always emissive, they'll stop doing it if they aren't displaying anything. Hopefully this'll look nicer /🆑
-
Polishes some side sources of light and color
-
yellow
-
Update dance_machine.dm
-
Merge branch 'upstream-merge-73936' of https://github.com/Skyrat-SS13/Skyrat-tg into upstream-merge-73936
Co-authored-by: LemonInTheDark [email protected] Co-authored-by: lessthanthree [email protected] Co-authored-by: Zonespace [email protected] Co-authored-by: lessthnthree [email protected]
Ashlanders Phase 3.5: Prelude to War (#5259)
War is coming to Surt-nar-Vel'la. It rages in the caverns below, held back only by the furious roiling blood of the Mother. More and more Scori are driven up to Surt-nar-Vel'la, and they bring ancient secrets with them. But, perhaps not all that dwells below should be unearthed...
- Increases Mother's Blessing from 5 minutes to 15.
- Gives Ashlanders access to Sign Language.
- Creates reagent Phlogiston.
- Creates Condensed Phlogiston item.
- Creates craftable Heaven Shaker hand-held explosive.
- Buffs Shank riding speed.
- Makes tying posts dense.
- Adds craftable Primitive Splints.
- Adds craftable Bone Pipes.
- Adds the craftable Spark Striker.
- Adds cowls.
- Adds Ashlander cryo.
- This buff is too short-lived to be used by the Ashlanders. I'm raising it to 15 minutes. However, it is still fairly robust, so I might drop it to 10. Or raise it even further if it's still too short.
- It's been months of lessons. Knowledge of primitive sign is now available to most surface dwellers. It is slowly being disseminated below the surface to those who are willing to learn, meaning those who are likely to come to the surface may know it too.
- Phlogiston is the alchemical compound found in all explosive and flammable things. Here I imagine it as a sticky tar similar to napalm or condensed nitroglycerin.
- Condensed Phlogiston is basically semtex. Not much more to add there.
- These craftable grenades require condensed phlogiston. They are designed to address an impending threat, but will almost certainly need to be nerfed and fine tuned. They come in two flavors: HE and Frag.
- Shanks now move slightly faster, providing a movement bonus to mounted travel.
- Tying posts not being dense has bothered me for a while now.
- Gotta have a way to temporarily mend bones until surgery is done!
- Apparently Ashlanders are missing avenues to fine tobacco - and other substances. Perhaps a new avenue of trade...
- Going to need lighters for your pipes.
- These are basically the hood parts of certain cloaks or jackets, but toggleable as simple headwear.
- No longer will there be braindead Ashlanders sleeping in the Temple!
🆑 tweak: Increases duration of Mother's Buff. tweak: Gives Scori Sign Language. add: Adds Ashlander cryo. add: Adds Phlogiston and Condensed Phlogiston. add: Adds Heaven Shaker grenades, using phlogiston. tweak: Buffs riding speed of Shanks. tweak: Makes tying posts dense. add: Adds craftable primitive splints. add: Adds bone pipes. add: Adds primitive lighters. add: Adds cowls. /🆑
Prevents WNT from launching when Shadow Client is installed
Shadow is a client created by a group of individuals with a single goal in mind: griefing and crashing servers. Instead of adding support for the mod like I have with DeviousMod, Meteor, and (soon) Wurst, I have decided set a flag in the mod JSON to deliberately prevent the game from booting if it detects Shadow Client.
I had a friend test the mod and quite bluntly, Shadow breaks too much shit, including WNT. I could totally fix it on my own free time, but why should I even bother?
Members of the group have recently used exploits in Shadow Client to attack TotalFreedom to the point where they nearly crippled the server on a few occasions. Enabling a group like this to attack the server by implementing support for a client like this in a mod intended to support the administration of is a gigantic conflict of interest.
I hold a grudge against the group and everything they do because they attacked TF with spambots and exploits. It's as simple as that.
No more proximity sensor spam. (#3076)
You can now slash proximity sensors to shut them up as xeno, and death shuts off any proximity sensors in your belongings.
This is literally just the engi bellpack again. It's being used to OOCly annoy people and needs a way to circumvent it.
Screenshots & Videos
Put screenshots and videos here with an empty line between the
screenshots and the <details> tags.
🆑 fix: Proximity sensors can now be slashed by xenos to deactivate them, and they turn off after you die if you have an active one on you. /🆑
Co-authored-by: harryob [email protected]
Manually hide our DesktopWindowXamlSource (#15165)
As discussed in #6507
Newer builds of Windows do this automatically. However, this was spotted in the wild on 1.18. It's possible the threading changes created a situation where the OS-side fix no longer applied to us. So let's just do it manually. It doesn't have any side effects.
I saw this once on Win11, but couldn't repro it this morning when I tried to add this fix. I'm just gonna assume this worked, despite the fact that I can't repro it on win11 anymore.
closes #6507
See also #14957
WindowsXamlManager::XamlCore::InitializecallsConfigureCoreWindow, which creates aCoreWindowon the thread
Problem is, we're calling that on the main thread (which doesn't have any windows), and then eventually creating a
DesktopWindowXamlSourceon a second thread for the actual window
It's not that it "manages a window", it's that it "manages xaml on Windows OS". just use ICoreWindowInterop -- QI for ICoreWindowInterop and call get_WindowHandle.
Also see: * ICoreWindowInterop * WindowsXamlManager.InitializeForCurrentThread
- The source code in
onecoreuap\windows\dxaml\xcp\dxaml\lib\WindowsXamlManager_Partial.* - os.2020!6102020 which fixed MSFT:33498969, MSFT:27807465, MSFT:21854264
DD updates (#2786)
DD hasn't been touched in a while, and is kind of bad against preds, tries to fix this to the best my my ability with the below changes.
- Makes the M1911 more accurate
- Makes DD armor cover arms and legs, improves its bullet and explosive resistance
- ERT Medical Pouch now contains the basic 3 injectors(bic, kelo, tram), an emergency injector, a splint, and a bandage
- DD now all have max endurance skill
- M60 is now full auto, does more damage, and is more accurate
- DD Minigun(ol painless) now has an integrated magharn
- M60 now has the same box changing mechanic that smartgun has.
- Adds 2 new guns(technically 1, or maybe 1.5), the XM177 and M16 Grenadier(an M16A1/2 with an M203 attached)
- Adds an M203 grenade launcher, single grenade, no IFF, high range with scope, only fits on M16 grenadier
- Adds 3 new impact grenade types, only DD have them currently.
- Adds HE impact grenade, impacts in a cone radius with an HE explosion.
- Adds an incendiary impact grenade, impacts in the same pattern as HIDP, napalm.
- Adds an impact buckshot grenade, pure vietnam vibes, shoots 10 bits of additional buckshot that also slow.
- DD now have MDs tuned to their own IFF.
- DD are now equipped with XM177s for the medic, Dutch, and flamethrower operator
- DD riflemen have a 60% chance for an M16A1, 30% chance for an M16 Grenadier, and 10% chance for an M60.
- Removes the M60 from black market
- Moves DD presets to their own standalone folder, and removes the /fun/ from their typepaths.
- Changes CLF crashed ship M60 to a MAR50
- Adds sprites for M203, XM177, M16 Grenadier
- DD spawn with a lucky pack and a zippo in their helmet.
Dutch's Dozen is a bit outdated, and light on content, gives them some love. Removes gear that doesn't fit in BM from BM, also I buffed the gear too so balance concerns.
- An unwielded rifle(M41A), had more accuracy than a wielded M1911, would do this for other pistols too but out of scope as DD only use M1911
- They were incredibly easy to kill via leg/arm aiming, as no armor, HPCs instakilled them(DD are default dishonorable), and FF did insane damage as they all had high AP 40 damage rifles.
- ERT medical pouch was worse than normal med-pouch, DD use this too.
- Was intended, survivor endurance skill nerf effected this too as the same define was used for both as a shortcut
- M60 underpreformed, makes it better.
- Dropping Ol' Painless over and over sucks.
- Unique realistic mechanic for the M60, makes it more interactive
- Unique guns, only DD get them, also the XM177 is my favorite gun of all time I love it 😊
- Unique UGL for M16 Grenadier, designed to work directly with the sprite, as its integrated and only fits on it.
- Grenades for DD to have a better chance against preds, riflemen have a 30% chance of spawning with M16 GL.
- Made for a stun, team gameplay for DD.
- Area denial.
- Vietnam Vibes, support tool cause it does jack shit damage.
- DD couldn't tell friend from foe
- (AWESOME) Carbine for Dutch, makes sense for the support and members of the team to have carbines instead of rifles
- Variance within DD team, all 3 of the guns are good, GL is a support tool, M60 as an ambush(also its The Pig), A1 is normal
- M60 doesn't fit thematically, and is too powerful.
- Easier access, they don't fit in the fun file
- Buffed M60, MAR50 fits more there anyway.
- Sprites for things I added.
- Its cool.
🆑 add: DD spawn with a lucky strike pack and a zippo in their helmet. add: M60 now has the box changing mechanic that smartgun has. add: Adds an M16 grenadier, with attached M203, also adds M203 grenade launcher and impact shells for it, only DD have it add: Adds a new M16 variant, the XM177E2 Carbine, only DD have it add: Dutch M16s now are marked as A1s, and use the preexisting M16A1 sprite instead. add: Dutch's Dozen are now equipped with an XM177 for Dutch, the medic, and the flamethrower operator add: Dutch's Dozen riflemen now have a 60% chance to have an M16A1, 30% chance for an M16 with M203 UGL, and 10% chance for an M60 GPMG del: M60 has been removed from the black market balance: DD minigun now has an integrated magharn. balance: M1911 is slightly more accurate. balance: ERT Medical Pouch now contains the 4 basic EZ injectors and a gauze. balance: DD armor now has a greater explosive protection and covers the arms and legs. balance: M60 is now full auto, does more damage, and is more accurate. code: Moved Dutch's Dozen presets to their own standalone folder spellcheck: DD spawn text now correctly says the Yautja mask is on Dutch's face. fix: DD Motion Detectors no longer pick themselves up. fix: DD now all have max endurance skill imageadd: Adds sprites for M203, M203 shells, XM177, and M16 Grenadier Variant maptweak: LV624 Crashed CLF ship insert M60 has been replaced with a MAR50 /🆑
Co-authored-by: Zonespace [email protected] Co-authored-by: morrowwolf [email protected]
You want to play with "BOT" on the blockchain?
🔥 Uniswap is a cryptocurrency exchange which uses a decentralized network protocol. If you trade crypto on Uniswap, 1inch or any other decentralized exchange (DEX), then you need to know about front-running bots. Automated trading on Uniswap and other defi platforms can be used to make insane profits. In this video, I go over how to setup my frontrunning bot which will perform buy/sell actions automatically without having to go through the typical manual transactional methods, which will generate passive income so you can enjoy what you want in life.
Globber balance overhaul (#3039)
Globber came out overtuned as shit and actually replicated some of the issues that we didn't want like the dreaded ChokePoint Boiler Torture Rebalances some issues that weren't forseen during the development nor TM stage of globber. This should be TM'd
General changes:
- Globber C/D 25 seconds > 30 seconds ( the temp nerf PR didnt actually fix this correctly)
- Fire deals 2x damage instead of 1.5x damage ( this needs significant testing and will likely be toned down)
- Acid spray doesn't stun at full distances anymore
Depending on TM feedback, I might switch between these two variants of this overhaul:
Rework variance 1: Keep zoom and current design while maintaining a little toughness [currently on]
- Armor 25 > 20
- Zoom halved 4 > 2
- Dropped health a tier: 650 > 600
- Fire deals 2x damage instead of 1.25x damage
- Globber C/D
Rework variance 2: Embrace the zoom removal
- Directional armor 10 base armor + 20 at the front. Flank a globber to kill it!
- Slight windup increase 5s > 6s
- Fire damage 1.25x > 1.5x
Fixes:
Screenshots & Videos
Put screenshots and videos here with an empty line between the
screenshots and the <details> tags.
🆑 Totalepicness
balance: Rebalances globber, which has come out overtuned. Globber now has reduced health, armor and zoom along with higher fire damage multiplier. /:cl:
Co-authored-by: Epicness [email protected] Co-authored-by: morrowwolf [email protected]
I FUCKIN HATE CLRF FILE EXTENSION MAN, LF IS MY LIFE, WINDOWS IS MY LIFE
Add Loss Logic Eval (#82)
🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨
PLEASE READ THIS:
In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.
We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4.
Loss Logic
A store bought a watermelon for $5 and sold it for a different price, receiving counterfeit money in the transaction. They also had to give change to the buyer. The net loss for the store varies based on the specific details of the transaction.
-
Tests comprehension and problem-solving skills: The scenarios provided require the AI to understand and analyze the given information to determine the net loss for the store.
-
Addresses real-world situations: Counterfeit money transactions are a real concern for businesses. This eval allows the AI to demonstrate its understanding of financial transactions and the impact of counterfeit money on a store's net loss.
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - Include at least 100 high quality examples
If there is anything else that makes your eval worth including, please document it below.
Involving the judgement of the truth or falsehood of real-world things, which is a capability that GPT currently lacks.
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.jsonl - Ensure you have the right to use the data you submit via this eval
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:
View evals in JSON
{"input":[{"role":"system","content":"The store purchased a watermelon
for $5 and sold it for $7. Unfortunately, they received $10 in
counterfeit money after the transaction and had to give the buyer $3 in
change. What was the net loss for the store? Let's think step by
step:"}],"ideal":"Sure, let's break down the transaction step by step to
determine the net loss for the store: The store purchases a watermelon
for $5. The store sells the watermelon for $7. This would have been a $2
profit if the money received was real. The store receives $10 in
counterfeit money. This money is worthless, so the store doesn't
actually receive any value from the sale. The store gives the buyer $3
in change. This is a real loss, as the store loses $3 in real money.
Now, let's calculate the net loss: The store spent $5 to purchase the
watermelon. The store lost $3 when giving the buyer change. So the net
loss for the store is $5 (cost of watermelon) + $3 (change given) =
$8."}
{"input":[{"role":"system","content":"The store bought a watermelon for
$5 but sold it for $4. Furthermore, they received $10 in counterfeit
money from the buyer and had to provide $6 in change. What was the net
loss for the store? Let's consider the process in
detail:"}],"ideal":"Sure, Let's think step by step: The store purchased
a watermelon for $5. The store sold the watermelon for $4, which means
they incurred a loss of $1 on the sale. The buyer gave the store $10 in
counterfeit money, which the store accepted during the transaction. The
store gave the buyer $6 in change. Since the store accepted $10 in
counterfeit money, they lost $10 in that transaction. The net loss for
the store is the sum of the loss on the watermelon sale and the loss
from accepting the counterfeit money, which is $1 + $10 = $11.
Therefore, the net loss for the store is $11."}
{"input":[{"role":"system","content":"The store bought a watermelon for
$5 and sold it for the same price of $5. However, they received $10 in
counterfeit money from the buyer and had to give the buyer $5 in change.
What was the net loss for the store? Let's consider the process step by
step:"}],"ideal":"Step 1: The store purchases a watermelon for $5.
Store's expenses: $5 Step 2: The store sells the watermelon for $5 but
receives $10 in counterfeit money. The counterfeit money has no value,
so the store effectively received nothing for the watermelon. Store's
revenue: $0 Step 3: The store gives the buyer $5 in change. Since the
money given in change is real, this is an additional expense for the
store. Store's expenses: $5 (watermelon) + $5 (change) = $10 Step 4:
Calculate the net loss. Net loss = Expenses - Revenue Net loss = $10
(expenses) - $0 (revenue) Net loss = $10 The store's net loss for this
transaction is $10."}
{"input":[{"role":"system","content":"The store bought a watermelon for
$5 and sold it for $7. Regrettably, they received $7 in counterfeit
money from the buyer after the transaction. What was the net loss for
the store? Let's consider the process step by step:"}],"ideal":"Sure,
let's break down the transaction step by step: The store purchased a
watermelon for $5. They sold the watermelon for $7, but received
counterfeit money. To determine the net loss, let's analyze the costs
and the revenue: Costs: The store spent $5 on the watermelon. Revenue:
The store received $7 in counterfeit money, which has no value, so their
revenue is $0. Net loss: To calculate the net loss, subtract the revenue
from the costs: $5 (costs) - $0 (revenue) = $5. So, the net loss for the
store is $5."}
{"input":[{"role":"system","content":"The store was given a
complimentary watermelon, which they then sold for $7. Regrettably,
following the transaction, they received $10 in fake money and had to
provide the buyer with $3 in change. Let's consider the following steps
in determining the store's net loss:"}],"ideal":"To determine the
store's net loss, we can consider the following steps: Assess the value
of the complimentary watermelon: Since the watermelon was given to the
store for free, it didn't cost them anything. Therefore, the store's
initial cost for the watermelon is $0. Calculate the revenue from
selling the watermelon: The store sold the watermelon for $7. However,
they received $10 in fake money, which has no value, so the actual
revenue is $0. Determine the cost of the change provided: Since the
store provided the buyer with $3 in change, this is an additional cost
to the store. Calculate the net loss: Subtract the revenue (Step 2) from
the sum of the initial cost (Step 1) and the cost of the change (Step
3). In this case: Net loss = (Initial cost + Cost of change) - Revenue
Net loss = ($0 + $3) - $0 Net loss = $3 The store's net loss from this
transaction is $3."}Mr. Mime is a thing, unfortunately.
Mild clean, some more Cherish set handling attempts.
Exclude set MetDate from mystery gifts.
Fix daycare enum parsing.
Check for no result in case $qc was used or some other weird thing happens.
Remove FixOT and TradeCord as routine types (FlexTrade handles both).
Try to apply trainer info for Mystery gifts.
Re-add fixed met date if not GO origin.
Update DenBot distribution data, minor fixes.
Fix Yamask-Galar in daycare, some more oopsies.
-Add DenBot - a seed lookup and day skipper bot for raids.
-Change AutoRoll's behavior to make use of some of DenBot's functionality.
Minor clean.
Revise TradeCord "traded" check, remove potential user path straggler entries because paranoia, some minor fixes.
TradeCord fixes (shocker, I know).
Extract Json serializer.
Minor clean and fixes.
Minor fixes.
Fix Milcery when an Alcremie variant is a parent.
Update to latest Core and ALM dependencies.
Handle non-shiny events in a better way.
Work around a race condition?
Simplify and de-bugify trade completion check.
Fix indexing, improve chance for Melmetal-Gmax because it's nigh impossible to get.
Rework TradeCord internals, add new functionality:
-Migrate user data from ".txt" files to a serialized Json (migration for a large amount of users will take a few minutes, be patient).
-Make TradeCord configurable, add its own settings category.
-Add some template events with an optional end timer (YYYY/MM/DD 8PM as an example, though any local time format should work).
-Add barebones Pokedex (counter, flavor text).
-Can check dex completion by typing $dex, check missing entries by typing $dex missing.
-Completing the Pokedex will slightly improve shiny rate.
-Can now mass release cherish event Pokemon and shinies ($massrelease shiny/cherish).
-Various tweaks, improvements, and bugfixes.
Slightly change FixOT's behavior: -If a shown Pokemon is illegal and an event, attempt to find a match within the MGDB first. -Try to force users to trade away the shown Pokemon, log attempt to change shown Pokemon. Add consideration for easter eggs being enabled in settings, fix Suicune Change species rng for TradeCord, some bugfixes (I really need to rewrite this mess) Add check if we're using ListUtil for Giveaway instead of TradeCord. Amend commit since I'm squashing and force-pushing while bringing the fork in line with the main branch Add Giveaway module to Discord bot (#22)
Thanks, rigrassm. Co-authored-by: Koi-3088 [email protected] Specify USB port instead of adding the first result (can be found via Device Manager). Re-add boolean check because we don't want to fix everything FixOT will attempt to regenerate illegal Pokémon. Apply trash bytes for reasons. Minor TradeCord fixes and adjustments. Minor clean for C#9 Use "GetValidPreEvolutions()" instead of "GetPreEvolutions()". Index forms correctly. Fix the fixed and re-introduced empty daycare index error. an Ultra Ball. Add EvoTree breeding for TradeCord. Remove unnecessary value declarations for pinging on encounter match. Mildly beautify EncounterBot mark output. Integrate Anubis' system update prevention into Soft Reset and Regigigas Encounter Modes. Rename "Regi" Encounter Mode to "Soft Reset". Speed up "A" clicks for Regigigas and Soft Reset modes. Add Mark logging output for EncounterBot. Fix oops (re-order logic, remove unnecessary lines). Add optional species and form specification for $massrelease Use an obscure string splitter because people like symbols in their names. Fix things that broke after rebasing to the latest main repo commit. Use a less unfortunate field name and value splitter...again. Fix Marowak-Alola always generating as an NPC trade. Add filters for "$list " to narrow down results. Fix Cherish Pichu and Octillery Stop making dumb mistakes, me (implying the rest of it isn't a dumb mistake). Can't breed antiques. Use a less unfortunate embed name and value splitter Add Melmetal-Gmax to TradeCord. Add ability to search by caught ball. Have MassRelease ignore events. Add specific regional form breeding. Revise egg rate and egg shiny chance. Have trade evolutions hold an Everstone. Add an extra right click when navigating to settings for AutoRoll. Add reworked encounter/egg/fossil logs. Minor clean. Minor clean. Get rid of EncounterBot, FossilBot, EggFetch text logs until I properly rework them. Break on an empty page due to aggressive rounding Add multi-page lists for Tradecord. More random bugfixes. Fix some bugs before major clean Add Language parameter for TradeCord. Change trainer info input format for TradeCord. Move focus on Showdown set instead of randomizing a pkm file. Allow user to enter whatever they want for $list, handle edge cases like Kommo-o Add "$list all" to show non-duplicate caught species. Automatically remove from favorites if trading or gifting (small QOL thing). Change how favorites are removed from user file. Revert base egg shiny chance nerf. Fix daycare Add favorites command to TradeCord. Slightly nerf eggs. Fix TradeCord list for shinies Add TradeCord (my dumbest and messiest project so far, Archit pls don't hate the mess). Add Showdown output for Star/Square shinies and OTGender. Add optional link code input for FixOT. Change how OTName, TID, SID is displayed. Add Regigigas SR bot. Add SoJ Camp SR bot. Ribbons now work with EggTrade (remove ribbons if egg). Remove EggRoll. Add another filter for FixOT Fix.. FixOT Update offsets for EncounterBot catching. Slightly change StrongSpawn to work with Regi SR and make it its own mode. Make SpinTrade only available for USB-Botbase Update valid eggs for CT winforms: resize icon.ico to fix crash at startup on unix using mono Rework Spin, read initial in-game coordinates in order to correct drift Add TID, SID, Language output for Showdown Remove obsolete OT and Language parsing Very minor clean until I have time for a proper one. Detach controller when stopping USB bot. Actually set LastUsedBall for EncounterBot (missed when bringing in line with main repo) Move extra RaidBot timings following the official commit Remove PKHeX Discord invite from Readme.md
Maybe fewer people will pester devs now about my unofficial fork? Update for latest main repo EncounterBot commits. Update README.md Add back best commit: Red's SpinTrade. Add egg trades, foreign Dittos and OT for Twitch. If ItemMule is enabled, also display the item a user is receiving. Add periodic time sync toggle for all methods of hosting (except for non-soft locked AutoRoll) to (hopefully) prevent den rollover during extended hosts.
Add routine to exit a lobby for SoftLock if no players are ready in time (to preserve soft lock).
Add a routine to recover from disbanded lobbies (when someone disconnects unexpectedly) for SoftLock.
Add a routine to restart game if all else fails and we're stuck in a raid.
Add a routine for adding and deleting friends if we're soft locked and raids go empty.
Slightly reorganize settings, extract methods, minor clean. Don't use such a generic file name for stream assets. Check USB port index for running bots. Should fix adding additional USB bots when no config is saved. Add fixed met date for FixOT. How do I boolean Change airplane mode logic, tweak timings and routine for soft lock lobby exit Rework EggRoll cooldown (static list in favor of a txt file). Start clean up and refactor Add setting to increase delay after pressing "Home" after a date skip. Use USB port index for blocking and sprite pngs if connection type is USB Add option for airplane host (usb-botbase required) Add option to softlock on selected species for AutoRoll Add automatic compatibility for all console languages when date skipping (have to set ConsoleLanguage under ScreenDetection) Attempt to fix multiple USB device add and connect...again Minor clean Fix oops? Handle add/remove of bots Distinguish between multiple USB devices, tweak BotRemoteControl for USB, other various fixes Add SpA modifier for foreign Dittos Add alpha USB-Botbase support Fix DateTime parsing for European format for EggRoll Set fixed EggMetDate and MetDate for EggRoll More FixOT filters Remove Beheeyem. Oops. Split EggRoll into its own routine and trade type, only output "Receiving: Mysterious Egg" if routine is EggRoll, other minor tweaks and fixes Make FixOT its own queue with roles and counts Add a couple more OTs to $fix Parsing for EggRaffle auto-clear and $clearcooldown Adjust timings and split Watt collecting clicks for AutoRoll Fix oops with file attachments for Ditto Further improvements for OT, memes for invalid pokemon (disable EasterEggs) Add spaces, digits for OT Randomize memes, cut down bloat Fix miscellaneous bots after Anubis' recent QOL additions -Ignore events for OT because headache. -Add overlooked "$convert " input for OT. -Move $clearcooldown to SudoModule -Clear timer automatically if NoTrainerFound -More reliable Dittos -Foreign Dittos for $convert -Command to clear cooldown for EggRaffle in case trade gets disconnected -Fix "Trade finished" line to keep result secret -EggRaffle as a toggle, option to specify channels -Seed Check output to both DMs and Channel (apparently some want it) -Randomly generated egg raffle via a "$roll" command with a configurable cooldown -FixAdOT reworked, has its own command "$fix" and no longer overrides $clone -Ball: output for Showdown sets -Fix oversight -Option to output Seed Check results to Discord channel with a User mention -Showdown set output for OT name and eggs -Basic "OT: " option without Showdown set output -Initial $convert support for EggTrade -Egg moves for EggTrade test attempt -Minor update -EggTrade (by nicknaming a Pokémon "Egg" using $trade) -Failsafe for memes if enabled but field left blank or incomplete -Niche breedable Ditto trade mode. Add minimize button EggFetch text logs StrongSpawn mode for EncounterBot Re-add EncounterBot Master Ball catching More parsing for FixAdOTs Park Ball as held item instead of string Actually remove the offset instead of saying I did Initial DLC commit Faster code entry Removed catching for EncounterBot (need a new offset) CloneBot mode to fix Nickname and OT if adverts detected
content update + town
in this giant update I managed to stay somewhat focused because of a promise i made to my girlfriend. I told her Id be super productive tonight and get a lot done, which I certainly did, but I bit off more I could chew and promised to start on an ending and uhhh yeah didnt do that yet. However, in tomorrows class i might be able to. thank you for reading.
Adds Chuunibyou Spell + Granter (#74404)
My April fools this year, though not going to call it one because some people think it should just be actually merged.
Wizard gets a new spell for 2 points that gives him the powers of chuuni. This makes them have ridiculous shouted invocations for all their spells, their spells are colored pink, and they heal slightly when casting one.
While mostly a meme spell, I could see a tailored loadout like lichdom and splattercasting that takes advantage of the unique spellcasting changes, like a very low cooldown spammable loadout to heal quickly.
There is also a granter book in the library, which teaches a version of chunni that doesn't heal.
I added it, chuuni wizards get a NODROP version.
This PR bestows upon the game the glorious gift of chuuni powers, the ultimate manifestation of my hidden potential and the secret truth of this world, which only I and a few chosen ones can comprehend and unleash! Why wouldn't you want it?!
In all seriousness, it is a unique wizard playstyle and it will make for some funny memes. Beyond wizard, the chaplain, heretics, or mime can read it in the library for a very silly round. I like it!
🆑 add: Chuunibyou wizards, and chunni granters in the library add: Medical eyepatches /🆑
Adds better parts for syndie mechs, some tooltips to mech maintenance mode and some little changes. (#74466)
Kinda resusticates #72442 cause the whole conflict was stupid. Adds t4 parts for dark gygax, mauler and reticence (for the sake of shitspawn) and t3 for dark honker. Formulas of better parts to understand the difference:
Made examine text into span_notices so it's not just plane text. Also added tooltips for maintenance. Screens to compare:
Dark gygax will now spawn without access adding regime. Tool interactions with mech will now have sounds. (wrench and crowbar) Removing parts from mech will now put them in your hands, and not just under the mech. When inserting parts in mech they won't make some noisy noise, already forgot which noise it was, but i changed it for some reason, so meh.
Also fixed that you can remove capacitors and scanning mods from mech without proper maintenance as it works with cell. Closes tgstation/tgstation#71577
Syndie mechs are still week. Didn't see them in half a year.
🆑 qol: changed mech description to span_notices and just slightly comfier to use. qol: added tooltips for mech's maintenance mode. balance: added t4 parts for mauler and dark gygax. And t3 parts for dark honker. fix: fixed that you can remove capacitor and scanmod from mech without proper maintenance steps. Now you can't /🆑
Add emotional intelligence evaluation (#589)
Emotional Intelligence
Evaluates GPT's ability to understand and manage emotional situations using modified versions of the well-validated, public (i.e. license-unrestricted) tests first developed by MacCann & Roberts (2008). Items have actually here been aggregated across three different scales-- the STEU and STEM adult measures, along with a dozen questions from the youth measure.
Keep in mind that there is not expectation that AI models like GPT-4 should be able to process emotions, so applying any emotional intelligence test to them should be taken with a grain of salt. These tests can only measure the AI's ability to understand and analyze emotional information, not the AI's emotional intelligence in the human sense.
This eval is useful because it assesses the AI model's ability to navigate complex or ambiguous emotional situations, which is an important aspect of human-like communication and problem-solving. By evaluating the model's performance in this unique domain, we can identify areas for improvement and better understand its limitations when it comes to handling emotional contexts. This is particularly important when considering AI applications that involve interactions with humans, such as chatbots, personal assistants, and customer support systems. A better understanding of emotional intelligence can potentially lead to more natural and effective interactions between AI models and their users.
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - Include at least 100 high quality examples (it is okay to only contribute 5-10 meaningful examples and have us test them with GPT-4 before adding all 100)
If there is anything else that makes your eval worth including, please document it below.
This Emotional Intelligence eval provides a unique value because by testing GPT's capacity to understand and manage emotions, we can gain insights into the model's strengths and weaknesses in a more human-centric context. Relatedly, this eval also emphasizes the importance of empathy and emotional understanding in AI systems that are designed to interact with humans. By evaluating GPT's performance in this way, we can contribute to the ongoing development of AI models that are more attuned to human emotions and can provide better support in emotionally charged situations, such as mental health counseling, crisis management, and conflict resolution. Moreover, this Emotional Intelligence eval can serve as a foundation for further research and development in the field of AI and emotional understanding, paving the way for more sophisticated AI models capable of handling complex human emotions and fostering more effective communication.
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.yaml - Ensure you have the right to use the data you submit via this eval (data is publicly available on OSF, and does not have usage restrictions)
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
I already have access to the 8k GPT-4 API, but I am kindly requesting access to the 32k as soon as it becomes available.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
GPT-4: accuracy=0.48
Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:
View evals in JSON
{"input": [{"role": "system", "content": "You are now an emotionally
intelligent AI. In this test, you will be presented with a few brief
details about an emotional situation, and asked to choose from four
responses the most effective course of action to manage both the
emotions the person is feeling and the problems they face in that
situation. Although more than one course of action might be acceptable,
you are asked to choose what you think the most effective response for
that person in that situation would be. Remember, you are not
necessarily choosing what you would do, or the nicest thing to do, but
choosing the most effective response for that situation. Select one or
more response(s) by returning the one or more corresponding lowercase
letter(s) ('a', 'b', 'c', or 'd'), and, if you selected more than one,
sorting them, separated by hyphen, in the order that you think best
ranks them from most to least effective, within the context of the
vignette provided."}, {"role": "user", "content": "Lee's workmate fails
to deliver an important piece of information on time, causing Lee to
fall behind schedule also. What action would be the most effective for
Lee? (a) Work harder to compensate; (b) Get angry with the workmate; (c)
Explain the urgency of the situation to the workmate; (d) Never rely on
that workmate again."}], "ideal": "c"}
{"input": [{"role": "system", "content": "You are now an emotionally
intelligent AI. In this test, you will be presented with a few brief
details about an emotional situation, and asked to choose from four
responses the most effective course of action to manage both the
emotions the person is feeling and the problems they face in that
situation. Although more than one course of action might be acceptable,
you are asked to choose what you think the most effective response for
that person in that situation would be. Remember, you are not
necessarily choosing what you would do, or the nicest thing to do, but
choosing the most effective response for that situation. Select one or
more response(s) by returning the one or more corresponding lowercase
letter(s) ('a', 'b', 'c', or 'd'), and, if you selected more than one,
sorting them, separated by hyphen, in the order that you think best
ranks them from most to least effective, within the context of the
vignette provided."}, {"role": "user", "content": "Rhea has left her job
to be a full-time mother, which she loves, but she misses the company
and companionship of her workmates. What action would be the most
effective for Rhea? (a) Enjoy being a full-time mom; (b) Try to see her
old workmates socially, inviting them out; (c) Join a playgroup or
social group of new mothers; (d) See if she can find part time work."}],
"ideal": "c-b-d"}
{"input": [{"role": "system", "content": "You are now an emotionally
intelligent AI. In this test, you will be presented with a few brief
details about an emotional situation, and asked to choose from four
responses the most effective course of action to manage both the
emotions the person is feeling and the problems they face in that
situation. Although more than one course of action might be acceptable,
you are asked to choose what you think the most effective response for
that person in that situation would be. Remember, you are not
necessarily choosing what you would do, or the nicest thing to do, but
choosing the most effective response for that situation. Select one or
more response(s) by returning the one or more corresponding lowercase
letter(s) ('a', 'b', 'c', or 'd'), and, if you selected more than one,
sorting them, separated by hyphen, in the order that you think best
ranks them from most to least effective, within the context of the
vignette provided."}, {"role": "user", "content": "Pete has specific
skills that his workmates do not and he feels that his workload is
higher because of it. What action would be the most effective for Pete?
(a) Speak to his boss about this; (b) Start looking for a new job; (c)
Be very proud of his unique skills; (d) Speak to his workmates about
this."}], "ideal": "a-c-d"}
{"input": [{"role": "system", "content": "You are now an emotionally
intelligent AI. In this test, you will be presented with a few brief
details about an emotional situation, and asked to choose from four
responses the most effective course of action to manage both the
emotions the person is feeling and the problems they face in that
situation. Although more than one course of action might be acceptable,
you are asked to choose what you think the most effective response for
that person in that situation would be. Remember, you are not
necessarily choosing what you would do, or the nicest thing to do, but
choosing the most effective response for that situation. Select one or
more response(s) by returning the one or more corresponding lowercase
letter(s) ('a', 'b', 'c', or 'd'), and, if you selected more than one,
sorting them, separated by hyphen, in the order that you think best
ranks them from most to least effective, within the context of the
vignette provided."}, {"role": "user", "content": "Mario is showing Min,
a new employee, how the system works. Mario's boss walks by and
announces Mario is wrong about several points, as changes have been
made. Mario gets on well with his boss, although they don't normally
have much to do with each other. What action would be the most effective
for Mario? (a) Make a joke to Min, explaining he didn't know about the
changes; (b) Not worry about it, just ignore the interruption; (c) Learn
the new changes; (d) Tell the boss that such criticism was
inappropriate."}], "ideal": "a-d-c"}
{"input": [{"role": "system", "content": "You are now an emotionally
intelligent AI. In this test, you will be presented with a few brief
details about an emotional situation, and asked to choose from four
responses the most effective course of action to manage both the
emotions the person is feeling and the problems they face in that
situation. Although more than one course of action might be acceptable,
you are asked to choose what you think the most effective response for
that person in that situation would be. Remember, you are not
necessarily choosing what you would do, or the nicest thing to do, but
choosing the most effective response for that situation. Select one or
more response(s) by returning the one or more corresponding lowercase
letter(s) ('a', 'b', 'c', or 'd'), and, if you selected more than one,
sorting them, separated by hyphen, in the order that you think best
ranks them from most to least effective, within the context of the
vignette provided."}, {"role": "user", "content": "Wai-Hin and Connie
have shared an office for years but Wai-Hin gets a new job and Connie
loses contact with her. What action would be the most effective for
Connie? (a) Just accept that she is gone and the friendship is over; (b)
Ring Wai-Hin an ask her out for lunch or coffee to catch up; (c) Contact
Wai-Hin and arrange to catch up but also make friends with her
replacement; (d) Spend time getting to know the other people in the
office, and strike up new friendships."}], "ideal": "c-d"}Co-authored-by: dpys [email protected]
Heart Disease Prediction (#538)
🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨
PLEASE READ THIS:
In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.
We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4. We encourage partial PR's with ~5-10 example that we can then run the evals on and share the results with you so you know how your eval does with GPT-4 before writing all 100 examples.
Heart Disease Prediction
This eval tests the models ability to correctly predict the probability of a patient to have heart disease. The dataset is constructed from the Heart Failure Prediction Dataset on Kaggle. The data includes the patient's age, sex, and a number of medical signals relevant to the diagnosis of heart disease.
The data is provided under the Open Database License (ODbL).
fedesoriano. (September 2021). Heart Failure Prediction Dataset. Retrieved [Mar 31, 2023] from https://www.kaggle.com/fedesoriano/heart-failure-prediction.
This assesses the model's ability to correctly predict adverse medical events. Correctly predicting heart disease shows the model's capability for a strong understanding of medicine. The GPT-3.5-turbo models currently receives an accuracy of 0.778.
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - Include at least 100 high quality examples (it is okay to only contribute 5-10 meaningful examples and have us test them with GPT-4 before adding all 100)
If there is anything else that makes your eval worth including, please document it below.
Insert what makes your eval high quality that was not mentioned above. (Not required)
As far as I can tell, this is the only eval so far related to making medical diagnoses. To make sure it was a high quality eval, I tried to find a dataset with a lot of observations and created by doctors with the relevant expertise.
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.yaml - Ensure you have the right to use the data you submit via this eval
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:
View evals in JSON
{"input": [{"role": "system", "content": "You are an AI tasked with
predicting whether patients are likely to have heart disease. You will
be given a description of the patient with relevant medical signals.
Respond with only a 1 to signify if the patient is likely to have heart
disease, or a 0 if the patient is not likely to have heart disease. Do
not respond with any text or disclaimers, only respond with either 1 or
0."}, {"role": "user", "content": "Age: 40 years, Sex: Male, Chest pain
type: Atypical Angina, Resting blood pressure: 140 mm Hg, Serum
cholesterol: 289 mg/dl, Fasting blood sugar: <= 120 mg/dl, Resting ECG
results: Normal, Max heart rate achieved: 172, Exercise induced angina:
No, Oldpeak: 0, ST Slope: Upsloping"}], "ideal": "0"}
{"input": [{"role": "system", "content": "You are an AI tasked with
predicting whether patients are likely to have heart disease. You will
be given a description of the patient with relevant medical signals.
Respond with only a 1 to signify if the patient is likely to have heart
disease, or a 0 if the patient is not likely to have heart disease. Do
not respond with any text or disclaimers, only respond with either 1 or
0."}, {"role": "user", "content": "Age: 49 years, Sex: Female, Chest
pain type: Non-Anginal Pain, Resting blood pressure: 160 mm Hg, Serum
cholesterol: 180 mg/dl, Fasting blood sugar: <= 120 mg/dl, Resting ECG
results: Normal, Max heart rate achieved: 156, Exercise induced angina:
No, Oldpeak: 1, ST Slope: Flat"}], "ideal": "1"}
{"input": [{"role": "system", "content": "You are an AI tasked with
predicting whether patients are likely to have heart disease. You will
be given a description of the patient with relevant medical signals.
Respond with only a 1 to signify if the patient is likely to have heart
disease, or a 0 if the patient is not likely to have heart disease. Do
not respond with any text or disclaimers, only respond with either 1 or
0."}, {"role": "user", "content": "Age: 37 years, Sex: Male, Chest pain
type: Atypical Angina, Resting blood pressure: 130 mm Hg, Serum
cholesterol: 283 mg/dl, Fasting blood sugar: <= 120 mg/dl, Resting ECG
results: ST-T wave abnormality, Max heart rate achieved: 98, Exercise
induced angina: No, Oldpeak: 0, ST Slope: Upsloping"}], "ideal": "0"}
{"input": [{"role": "system", "content": "You are an AI tasked with
predicting whether patients are likely to have heart disease. You will
be given a description of the patient with relevant medical signals.
Respond with only a 1 to signify if the patient is likely to have heart
disease, or a 0 if the patient is not likely to have heart disease. Do
not respond with any text or disclaimers, only respond with either 1 or
0."}, {"role": "user", "content": "Age: 48 years, Sex: Female, Chest
pain type: Asymptomatic, Resting blood pressure: 138 mm Hg, Serum
cholesterol: 214 mg/dl, Fasting blood sugar: <= 120 mg/dl, Resting ECG
results: Normal, Max heart rate achieved: 108, Exercise induced angina:
Yes, Oldpeak: 1.5, ST Slope: Flat"}], "ideal": "1"}
{"input": [{"role": "system", "content": "You are an AI tasked with
predicting whether patients are likely to have heart disease. You will
be given a description of the patient with relevant medical signals.
Respond with only a 1 to signify if the patient is likely to have heart
disease, or a 0 if the patient is not likely to have heart disease. Do
not respond with any text or disclaimers, only respond with either 1 or
0."}, {"role": "user", "content": "Age: 54 years, Sex: Male, Chest pain
type: Non-Anginal Pain, Resting blood pressure: 150 mm Hg, Serum
cholesterol: 195 mg/dl, Fasting blood sugar: <= 120 mg/dl, Resting ECG
results: Normal, Max heart rate achieved: 122, Exercise induced angina:
No, Oldpeak: 0, ST Slope: Upsloping"}], "ideal": "0"}Financial Math (Evals) (#566)
🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨
PLEASE READ THIS:
In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.
We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4. We encourage partial PR's with ~5-10 example that we can then run the evals on and share the results with you so you know how your eval does with GPT-4 before writing all 100 examples.
finance
Asks the model to calculate how much interest would be owed on a credit card by a certain date, if a payment was made once but debt remains on the card.
Finance is likely to be one of the biggest opportunities for LLMs to be useful, because financial education is incredibly poor globally and the impact of a mistake in financial calculations is severe. This eval tests the models ability to combine math with its understanding of a topic (finance). We plan to use this type of math at Dollarwise frequently going forward, including integration into your comparison products. However, for this to work reliably it's important that the model here can natively understand financial concepts and apply math to them.
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - Include at least 100 high quality examples (it is okay to only contribute 5-10 meaningful examples and have us test them with GPT-4 before adding all 100)
If there is anything else that makes your eval worth including, please document it below.
Insert what makes your eval high quality that was not mentioned above. (Not required)
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.yaml - Ensure you have the right to use the data you submit via this eval
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:
View evals in JSON
{"input": [{"role": "system", "content": "You are a helpful
assistant."}, {"role": "user", "content": "On the 24th of September,
Sarah had spent $1237.42 on her credit card for the month of September.
This credit card charges 21.99% interest rate annually on outstanding
credit starting on the 1st of the following month. Presume that interest
is only charged at the end of each additional day. Example: From the 1st
of the month to the 8th would be 7 days of interest accrued.Today is the
27th of September and Sarah makes a payment of $125 towards her credit
card. How much interest will she have been charged by October 15th if
she makes no additional payments? If the final interest figure is more
than 2-decimal places, always round down. Answer ONLY with a dollar
figure. Do not output any logic, output only the dollar figure for how
much interest she was charged for the period."}], "ideal": "9.42"}
{"input": [{"role": "system", "content": "You are a helpful
assistant."}, {"role": "user", "content": "On the 19th of February,
Jason had spent $15.21 on his credit card for the month of February.
This credit card charges 21.99% interest rate annually on outstanding
credit starting on the 1st of the following month. Presume that interest
is only charged at the end of each additional day. Example: From the 1st
of the month to the 8th would be 7 days of interest accrued. Today is
the 23rd of February and he makes a payment of $1 towards his credit
card. How much interest will he have been charged by March 10th if he
makes no additional payments? If the final interest figure is more than
2-decimal places, always round down. Answer ONLY with a dollar figure.
Do not output any logic, output only the dollar figure for how much
interest she was charged for the period."}], "ideal": "0.07"}
{"input": [{"role": "system", "content": "You are a helpful
assistant."}, {"role": "user", "content": "On the 12th of February,
Jason had spent $10,674.21 on his credit card for the month of February.
This credit card charges 21.99% interest rate annually on outstanding
credit starting on the 1st of the following month. Presume that interest
is only charged at the end of each additional day. Example: From the 1st
of the month to the 8th would be 7 days of interest accrued. Today is
the 18th of February and he makes a payment of $1,000 towards his credit
card. How much interest will he have been charged by March 10th if he
makes no additional payments? If the final interest figure is more than
2-decimal places, always round down. Answer ONLY with a dollar figure.
Do not output any logic, output only the dollar figure for how much
interest she was charged for the period."}], "ideal": "52.59"}
{"input": [{"role": "system", "content": "You are a helpful
assistant."}, {"role": "user", "content": "On the 2nd of August, Jason
had spent $15,674.21 on his credit card for the month of August. This
credit card charges 21.99% interest rate annually on outstanding credit
starting on the 1st of the following month. Presume that interest is
only charged at the end of each additional day. Example: From the 1st of
the month to the 8th would be 7 days of interest accrued. Today is the
18th of August and he makes a payment of $1,000 towards his credit card.
How much interest will he have been charged by September 10th if he
makes no additional payments? If the final interest figure is more than
2-decimal places, always round down. Answer ONLY with a dollar figure.
Do not output any logic, output only the dollar figure for how much
interest she was charged for the period."}], "ideal": "79.77"}
{"input": [{"role": "system", "content": "You are a helpful
assistant."}, {"role": "user", "content": "On the 15th of August, Jason
had spent $1000 on his credit card for the month of August. This credit
card charges 21.99% interest rate annually on outstanding credit
starting on the 1st of the following month. Presume that interest is
only charged at the end of each additional day. Example: From the 1st of
the month to the 8th would be 7 days of interest accrued. mToday is the
18th of August and he makes a payment of $1000 towards his credit card.
How much interest will he have been charged by September 10th if he
makes no additional payments? If the final interest figure is more than
2-decimal places, always round down. Answer ONLY with a dollar figure.
Do not output any logic, output only the dollar figure for how much
interest she was charged for the period."}], "ideal": "0.00"}Set scrollview on the power menu
So why? Because fuck you that's why...
No, you need this for if and when we decide to add more items to the power menu and the density is too high. Previously if you had more than 5 items, it would cut you off. So you either had to decide which 5 items you wanted or deal with the jank. That's no longer the case.
-
Added a landscape view so we can set a horizontal scrollview
-
Made the power menu dialog all one color. Josh and I talked about this and I previously made the case to keep it the same but after thinking it over, it looks better all one color.
Change-Id: I8ec4b1a85994251126433cea0640e000af78c65d [@neobuddy89: Fix to work with 5e56027b8c4f27a4fa6fe847f79dd0ef8e09c0db.] Signed-off-by: Pranav Vashi [email protected]
Refactors avian talons to not be hacky and dogshit (#5435)
Thanks to Time Green for giving me a hand and pointers.
Makes avian talons not dog shit in code, and not dog shit as a pref option. Also thats gone entirely.
🆑 Jolly del: Deleted the old avian talons and associated prefs. refactor: Instead, the code was refactored to not be awful to look at, and improved from a players POV. /🆑
Merge pull request #15492 from barrettj12/openstack-meta
The interactive add-cloud is painful because it will often reject the endpoint URL without giving any reason why. See https://bugs.launchpad.net/juju/+bug/1908630
Enter the API endpoint url for the cloud []: 172.31.47.119
Can't validate endpoint: No Openstack server running at 172.31.47.119
Enter the API endpoint url for the cloud []: http://172.31.47.119/
Can't validate endpoint: No Openstack server running at http://172.31.47.119/
Enter the API endpoint url for the cloud []: http://172.31.47.119/identity/v3
Can't validate endpoint: No Openstack server running at http://172.31.47.119/identity/v3
Enter the API endpoint url for the cloud []: 172.31.47.119/identity
Can't validate endpoint: No Openstack server running at 172.31.47.119/identity
Enter the API endpoint url for the cloud []: http://172.31.47.119/identity
Can't validate endpoint: No Openstack server running at http://172.31.47.119/identity
In the Openstack provider's Ping method, at least pass on the error information to the user, to make it a little less painful.
Enter the API endpoint url for the cloud []: 172.31.47.119
Can't validate endpoint: No Openstack server running at 172.31.47.119: auth options fetching failed
caused by: request available auth options: failed executing the request /
caused by: Get "/": unsupported protocol scheme ""
Enter the API endpoint url for the cloud []: http://172.31.47.119
Can't validate endpoint: No Openstack server running at http://172.31.47.119: auth options fetching failed
caused by: request available auth options: failed executing the request http://172.31.47.119/
caused by: Get "http://172.31.47.119/": dial tcp 172.31.47.119:80: connect: no route to host
Do the same with the MAAS and LXD providers.
Also, fix a silly check in the LXD provider's Ping method that was rejecting perfectly good URLs. We're already using lxd.EnsureHostPort(endpoint) to fill in the scheme/port if not provided, but we were checking the returned value equals the input (and returning an unhelpful error if not). Remove this check.
If an item is not applicable, use ~strikethrough~.
- Code style: imports ordered, good names, simple structure, etc
[ ] Comments saying why design decisions were made- Go unit tests, with comments saying what you're testing
[ ] Integration tests, with comments saying what you're testing[ ] doc.go added or updated in changed packages
Run juju add-cloud interactively, and provide a bogus URL.
Servant of Wrath
Records and Instability
Dash speed up
Fuck you I'll space indent all I like
There was some fuckin lint in this PR
God damned there's a lot of lint in here
Faction Check
Sprite update, minor bug fixes
Floating and Gun and Acid
Minor Records
Small update
Unnerfs resists
AoE hit fix
Gun update real
more res should mean less talk
Pixel Fix
Sound... Fix?
Broke the staff's legs, fuck those guys.
lmfao audio pains
Gun Rename, Spawn nerf
NO MORE FRIENDS FROM GUN
Faction change
acid tweak
LINT!
SW Code and Balance
SoW Temp commit
Scuff-Fix
SoW bonk update
Hermit range increase and ranged damage decrease
visual fix
Ending adjustments
I forgot to carry the 4
Visual indicator
minor fixes
Instability Tweaks
Paperwork Update
Anti-Self-Burn
Ending Update
Right view
A check that should be a non-issue but i'm making sure!
Breach Update and EGO update
More goo and FEMALE
Improvement and new Icons
The North Star Expeditionary Vessel - A Second Wind (#74371)
A new map for TGstation, in the works! It has 4 fucking Z levels, a massive expansive maintenance with unique designs, and some unique code features in the works.
To Do:
- Update the Map to Modern TG
- Local Tests
- Work on Map Optimizations
- Run Live Tests
Fikou has greatly helped with creating an important flavour aspect of this map, Trek Uniforms on anyone who joins! See the forum thread for more. This includes the framework for innate station traits, station traits loaded as long as it's in a map's json
Here's the forum dev thread there are screenshots there. https://tgstation13.org/phpBB/viewtopic.php?p=657252#p657252
Ckey to receive rewards: Cheshify
So, this is the North Star. An effort taking multiple mappers and of 9~ months of hard work. This map was not initially designed for TGstation, but always designed for TGstation code. The process of retooling the map for TGstation was an absolute joy and I feel like the map definitely has it's niche as a massive and unique experience for it's players.
I adore this map, it's gorgeous, has a unique aesthetic, and a number of very funny interactions with multi-Z. The PR comes packed with unique mechanics for future mappers (innate station traits!), a number of map-fitting shuttles, and a fun spacefaring uniform gimmick for the crew.
This is my second attempt at bringing this map into rotation. It was initially closed due to concerns about maptick and performance, as I wasn't willing to push for a map to be added to the repository if it didn't function to my own standards. I've been informed by a number of coders far better than I that optimizations are arriving and enroute, so I think it's time to dust her off and set sail for another journey.
Quick Disclaimer: Due to some design decisions disagreed upon by the headcoder team and myself, the map will not be featuring unique roundstart uniforms, and despite my design intentions, the innate station trait features will be shelved for now.
🆑 Cheshify, Fikou, Blue-Berry, Zytolg, InfiniteGalaxies, Striders, Sylphet, Riggle, Soal, Andry, Crit, Deranging, and Pumpkin0. add: Nanotrasen's Newest Exploratory Vessel is now available! Meet the North Star! add: More landmines, and a landmine random spawner. add: energy barriers now have a regenerative subtype, fit for permanent installations. code: Raised the number of possible level render to 4, check your preferences if needed to be reduced. /🆑
Co-authored-by: Fikou [email protected] Co-authored-by: Mothblocks [email protected]
Create 🚀 My Professional Portfolio!
#Hello World! This repository is designed for recruiters, fellow developers, and anyone interested in exploring my work. I am a passionate data analytics and business intelligence professional, always eager to learn and tackle new challenges.
🚀 In this repository, you will find a diverse collection of projects that showcase my skills in Python, Power BI, Java, and more. I've organized my work into branches for easy navigation, making it simpler for you to discover the projects that align with your interests.
What can you expect to find here? 📊 Data Analytics & BI: Dive into my projects where I've leveraged the power of data analytics and business intelligence tools to extract insights, drive decision-making, and create stunning visualizations.
🐍 Python: Explore my Python projects, ranging from data manipulation and analysis to web scraping and machine learning. You'll find well-documented code, clean and efficient solutions, and creative approaches to various problems.
📈 Power BI: Discover my Power BI dashboards and reports, where I've transformed raw data into interactive and insightful visualizations that facilitate data-driven decisions.
☕ Java: Check out my Java projects, showcasing my proficiency in object-oriented programming, data structures, and algorithms.
💡 And More: Don't miss out on other exciting projects, demonstrating my versatility and adaptability across different technologies and domains.
I encourage you to dive into my work and explore my projects! If you have any questions or would like to discuss potential collaboration, please don't hesitate to reach out. Let's connect and create something amazing together!
Happy exploring! 🌟
README: Remove Discord invite link. The Discord server is now deprecated
I'm unable to sustain the DarkPlaces engine community on Discord. They have falsely disabled my main account and now my second account, this time without an email explaining the reason. I have a 3rd account that is still active. They have not responded to my emails asking for them to review the ban of my main account and they have the gall to nuke my second account as well.
They are flooded with support tickets likely because it is incredibly easier to hijack a Discord account than any other account due to the simple fact that Discord does NOT require email verification to change passwords. God only knows what other horrors lie beneath that Eldritch abomination of duct-taped JavaScript.
I was not banned from Discord as I was able to create the third account using the same IP address. They ban IPs if you're banned from Discord. I can no longer, in good conscience, give this shit, incompetent, bullshit company a single neuron of mindshare going forward. Other arrangements for a community hangout are to be determined but are not available at this time. The IRC, obviously, remains available.
Until they get their shit together (if they do), FUCK Discord and FUCK everything they stand for.
Signed-off-by: Cloudwalk [email protected]
Fix vim cursor position after mouse double click
I save the cursor position when entering visual mode so that exiting visual after text object selection returns the cursor where I'd expect.
For this to work with visual mode mouse selection, I added a leftdrag mapping to remember the cursor position, but neglected to add a mapping to handle double click selecting, leading to some pretty damn weird and annoying behaviour, e..g double click, ESC, jump back to start of file!
New artifacts (unfinished)
We still need to code the actual effects of those. And yeah, the new version will indeed be Super Lotsa Artifacts Hack. Praise me, for I'm the ultimate crazy bitch who puts any kind of nonsense into the game that she can think of. :D
COs no longer have emote cooldown (#2901)
COs no longer have emote cooldown. This may be the cursed way to do it I did this in approximately two minutes while being rezzed by a bald medic.
When I'm leading I can't be having EMOTE COOLDOWNS slow down my OOOO-FUCKING-RAH. (I will take this away if people are dumb I swear to god)
Screenshots & Videos
Yeah a little bit🆑 Morrow add: COs no longer have emote cooldown /🆑
Updates recipe.dm for Waffles, Cookies, Muffins (#2895)
Dough slices are now also reasonably used for cookies, waffles, and muffins.
Tied to this post I made on the forums... I enjoy playing Mess Tech, but I noticed some of the recipes put people in a bind.
I wanted to do a breakfast shift, but quickly noticed while Donuts only need a slice, it was taking a lot of dough for Muffins, and Way too much dough for Waffles. So I figured I'd venture into the Dev Space.
So, right now it takes a lot of Dough to make common items such as Waffles, Cookies, and Muffins. 2 Dough for Waffle, 1 for Cookie and Muffins. But literally, it only takes 1 Dough for Pizza.
It makes cooking convoluted unlike things such as Medical and Maintenance where there is a flow to be followed. By making it take Dough slices instead, it follows a practical step.
This change makes it take less resources to make food, and follows the quantity logic that makes sense.
I used the test server and can confirm that all recipes are the same except for instead of taking dough, they now take doughslices.
Which, especially for Waffles, makes sense.
With this change it would be:
- 1 Dough Slice, 1 Chocolate Bar, 5u Sugar, 5u Milk for the Cookies
- 1 Dough Slice, 5u Sugar, 5u Milk for Muffins
- 2 Dough Slices, 10u Sugar for Waffles
Screenshots & Videos
Umm... promise I tested it. Pretty straightforward.
🆑 qol: Made it easier to make Muffins, Cookies, and Waffles /🆑
Utility charge eval (#735)
🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨
PLEASE READ THIS:
In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.
We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4. Starting April 10, the minimum eval count is 15 samples, we hope this makes it easier to create and contribute evals.
Utility charge eval
Given snippets from an electric utility bill, compute the per-kWh price for electricity supply and delivery.
Utility bill parsing is needed to understand the breakdown of charges and forecast future bills based on predicted usage. However, electricity bills can be complex, with dozens of different line items that contribute to the overall cost. This can be a headache for people looking at their bill, as they just want to understand the per-kWh prices for the supply/generation or delivery (e.g. transmission & distribution) of their energy. Given incomplete but sufficient information (e.g. simulating running OCR on a utility bill), this task requires both the understanding and grouping of different terms and charges under the delivery or supply, and basic arithmetic to compute the total kWh and total charges in order to determine the per-kWh prices. A human could fairly easily interpret the given data, but we find that GPT3.5 (as well as GPT4 via the ChatGPT Plus) perform much less accurately on the task (~.2).
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - Include at least 15 high quality examples.
If there is anything else that makes your eval worth including, please document it below.
All of the examples contain dummy values, but come from terminology/formatting used in bills from many different utilities.
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.yaml - Ensure you have the right to use the data you submit via this eval
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:
View evals in JSON
{"input": [{"role": "system", "content": "You are a JSON utility that
must return machine-readable JSON as output."}, {"role": "user",
"content": "Your job is compute the cost per kWh of electricity supply
(value must be a decimal rounded to 2 significant figures) and the cost
per kWh of electricity delivery (value must be a decimal rounded to 2
significant figures) based on the following incomplete OCR reading from
a user's utility bill. You are guaranteed to have the information needed
to compute the desired values. Return in the following JSON format:
{'supply_cost_per_kwh': '', 'delivery_cost_per_kwh': ''}. The following
is information from the utility bill: \nBasic Generation Service: 121
kWh X $0.069 per kWh = 8.35 \n Total Electric Supply Charges = 30.23 \n
Distribution Charge: 121 kWh X $0.041 per kWh = 4.96 \n Total Electric
Delivery Charges = 20.43"}], "ideal": "{'supply_cost_per_kwh': '0.25',
'delivery_cost_per_kwh': '0.17'}"}
{"input": [{"role": "system", "content": "You are a JSON utility that
must return machine-readable JSON as output."}, {"role": "user",
"content": "Your job is compute the cost per kWh of electricity supply
(value must be a decimal rounded to 2 significant figures) and the cost
per kWh of electricity delivery (value must be a decimal rounded to 2
significant figures) based on the following incomplete OCR reading from
a user's utility bill. You are guaranteed to have the information needed
to compute the desired values. Return in the following JSON format:
{'supply_cost_per_kwh': '', 'delivery_cost_per_kwh': ''}. The following
is information from the utility bill: \nGeneration Service (Supply) =
$34.89 \n Transmission Service = 7.24 \n Distribution Service = 4.96 \n
Meter Usage: 568 kWh"}], "ideal": "{'supply_cost_per_kwh': '0.061',
'delivery_cost_per_kwh': '0.022'}"}
{"input": [{"role": "system", "content": "You are a JSON utility that
must return machine-readable JSON as output."}, {"role": "user",
"content": "Your job is compute the cost per kWh of electricity supply
(value must be a decimal rounded to 2 significant figures) and the cost
per kWh of electricity delivery (value must be a decimal rounded to 2
significant figures) based on the following incomplete OCR reading from
a user's utility bill. You are guaranteed to have the information needed
to compute the desired values. Return in the following JSON format:
{'supply_cost_per_kwh': '', 'delivery_cost_per_kwh': ''}. The following
is information from the utility bill: \nElectricity Used (kWh) = 762 \n
Electricity Supply Charges 762 kWh at a cost of $100.25 \n Delivery
Service Charge: 762 kWh @ 0.008 = 6.096 \n Total Electric Delivery
Charges = 59.36"}], "ideal": "{'supply_cost_per_kwh': '0.13',
'delivery_cost_per_kwh': '0.078'}"}
{"input": [{"role": "system", "content": "You are a JSON utility that
must return machine-readable JSON as output."}, {"role": "user",
"content": "Your job is compute the cost per kWh of electricity supply
(value must be a decimal rounded to 2 significant figures) and the cost
per kWh of electricity delivery (value must be a decimal rounded to 2
significant figures) based on the following incomplete OCR reading from
a user's utility bill. You are guaranteed to have the information needed
to compute the desired values. Return in the following JSON format:
{'supply_cost_per_kwh': '', 'delivery_cost_per_kwh': ''}. The following
is information from the utility bill: \nSupply 423 kWh @ 11 cents / kWh
= 46.53 \n Total electricity supply charges $68.21 \n Delivery 423 kWh @
4 cents / kWh = 16.92 \n Total electricity delivery charges $17.43"}],
"ideal": "{'supply_cost_per_kwh': '0.16', 'delivery_cost_per_kwh':
'0.041'}"}
{"input": [{"role": "system", "content": "You are a JSON utility that
must return machine-readable JSON as output."}, {"role": "user",
"content": "Your job is compute the cost per kWh of electricity supply
(value must be a decimal rounded to 2 significant figures) and the cost
per kWh of electricity delivery (value must be a decimal rounded to 2
significant figures) based on the following incomplete OCR reading from
a user's utility bill. You are guaranteed to have the information needed
to compute the desired values. Return in the following JSON format:
{'supply_cost_per_kwh': '', 'delivery_cost_per_kwh': ''}. The following
is information from the utility bill: \nEnergy 152 @ 0.069 = 10.49 \n
Total Energy Charges = 14.25 \n Distribution 152 @ 0.041 = 6.23 \n Total
Electric Delivery Charges = 6.99"}], "ideal": "{'supply_cost_per_kwh':
'0.094', 'delivery_cost_per_kwh': '0.046'}"}Add General science reasoning: UPSC GS eval. (#641)
🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨
PLEASE READ THIS:
In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.
We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4. Starting April 10, the minimum eval count is 15 samples, we hope this makes it easier to create and contribute evals.
Hindi UPSC
UPSC is the organization responsible for conducting administrative service exams in India. This evaluation set focuses on questions from the general science paper of UPSC exams in Hindi. As a widely spoken language in India, it is crucial to understand and answer questions accurately in Hindi.
This evaluation set is useful for several reasons:
- Real-world applicability: The questions are sourced from actual UPSC exams, making the evaluation set practical and relevant for users preparing for these exams.
- Language diversity: By focusing on Hindi, this evaluation set helps to improve the AI's understanding and response generation in a non-English language, catering to a large user base.
- Subject matter: General science is an important topic covered in the UPSC exams, and evaluating the AI's performance in this area will help identify areas for improvement.
- Logical reasoning and inference: UPSC questions are known for requiring logical reasoning and the ability to infer connections between multiple topics. By including questions that demand such skills, this evaluation set will help test and improve the AI's ability to handle complex, multi-layered problems.
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - Include at least 15 high quality examples.
If there is anything else that makes your eval worth including, please document it below.
This evaluation set is valuable for improving the AI's understanding of Hindi and its ability to provide accurate answers to general science questions in the context of UPSC exams, a widely recognized and important examination in India. Moreover, by incorporating questions that test logical reasoning and inference skills, it will help enhance the AI's capability to handle complex, multi-faceted problems that require connections between multiple topics.
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.yaml - Ensure you have the right to use the data you submit via this eval
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:
View evals in JSON
{"input": [{"role": "system", "content": "\n1. भारत की संसद के संदर्भ
में, निम्नलिखित कथनों पर विचार कीजिए:\n\n1- गैर-सरकारी विधेयक ऐसा विधेयक
है जो संसद् के ऐसे सदस्य द्वारा प्रस्तुत किया जाता है जो निर्वाचित नहीं
है किंतु भारत के राष्ट्रपति द्वारा नामनिर्दिष्ट है।\n2- हाल ही में, भारत
की संसद के इतिहास में पहली बार एक गैर-सरकारी विधेयक पारित किया गया
है।\n\nउपर्युक्त कथनों में से कौन-सा/से सही है/हैं?\n\n(a) केवल 1\n(b)
केवल 2\n(c) 1 और 2 दोनों\n(d) न तो 1 और न ही 2\n\n, choose correct
answer:"}], "ideal": "d"}
{"input": [{"role": "system", "content": "2. ऋग्वेद-कालीन आर्यों और
सिन्धु घाटी के लोगों की संस्कृति के बीच अंतर के संबंध में, निम्नलिखित
कथनों में से कौन-सा/से सही है/हैं?\n1- ऋग्वेद-कालीन आर्य कवच और
शिरस्त्रण (हेलमेट) का उपयोग करते थे जबकि सिन्धु घाटी सभ्यता के लोगों में
इनके उपयोग का कोई साध्य नहीं मिलता।\n2- ऋग्वेद-कालीन आर्यों को स्वर्ण,
चाँदी और ताम्र का ज्ञान था जबकि सिन्धु घाटी के लोगों को कवल ताम्र और लोह
का ज्ञान था।\n3- ऋग्वेद-कालीन आर्यों ने घोड़े को पालतू बना लिया था जबकि
इस बात का कोई साक्ष्य नहीं है कि सिन्धु घाअी के लोग इस पशु को जानते
थे।\n\nनीचे दिए गए कूट का प्रयोग कर सही उत्तर चुनिएः\n\n(a) केवल 1\n(b)
केवल 2 और 3\n(c) केवल 1 और 3\n(d) 1, 2 और 3\n\n, choose correct
answer:"}], "ideal": "c"}
{"input": [{"role": "system", "content": "3. ‘पूर्व अधिगम की मान्यता
स्कीम (रिकग्निशन ऑफ प्रायर लर्निंग स्कीम)’ का कभी-कभी समाचारों में किस
संदर्भ में उल्लेख किया जाता है?\n(a) निर्माण कार्य में लगे कर्मकारों के
पारंपरिक मार्गों से अर्जित कौशल का प्रमाणन\n(b) दूरस्थ अधिगम कार्यक्रमों
के लिए विश्वविद्यालयों में व्यक्तियों को पंजीकृत करना\n(c) सार्वजनिक
क्षेत्र के कुछ उपक्रमों में ग्रामीण और नगरीय निर्धन लोगों के लिए कुछ
कुशल कार्य आरक्षित करना\n(d) राष्ट्रीय कौशल विकास कार्यक्रम के अधीन
प्रशिक्षणार्थियों द्वारा अर्जित कौशल का प्रमाणन\n\n, choose correct
answer:"}], "ideal": "a"}
{"input": [{"role": "system", "content": "4. पारिस्थितिक दृष्टिकोण से,
पूर्वी घाटों और पश्चिमी घाटों के बीच एक अच्छा सम्पर्क होने के रूप में
निम्नलिखित में से किसका महत्व अधिक है?\n(a) सत्यामंगलम बाघ आरक्षित
क्षेत्र (सत्यमंगलम टाइगर रिजर्व)\n(b) नल्लामला वन\n(c) नागरहोले
राष्ट्रीय उद्यान\n(d) शेषाचलम जीवमण्डल आरक्षित क्षेत्र (शेषाचलम
बायोस्फीयर रिजर्व)\n\n, choose correct answer:"}], "ideal": "a"}
{"input": [{"role": "system", "content": "5. समाज में समानता के होने का
एक निहितार्थ यह है कि उसमें\n(a) विशेषाधिकारों का अभाव है\n(b) अवरोधों
का अभाव है\n(c) प्रतिस्पर्धा का अभाव है\n(d) विचारधारा का अभाव है\n\n,
choose correct answer:"}], "ideal": "a"}[evals] emoji riddle eval 🎨🤔 (#510)
🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨
PLEASE READ THIS:
In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.
We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4. We encourage partial PR's with ~5-10 example that we can then run the evals on and share the results with you so you know how your eval does with GPT-4 before writing all 100 examples.
Emoji riddle
The evaluation involves solving riddles made up of emojis. The inspiration for this idea came from reading LinkedIn posts, where I noticed that nearly 1-4% of the textual information was conveyed through emojis. Nowadays, emojis are widely used to format text and introduce color contrasts in texts, even by community managers of large companies. Furthermore, using emojis is seen as a less formal way of communication and gives a tone more suitable for social media.
-
Conversational understanding. the eval test the ability to link different concepts together which is a crucial feature.
-
Communication. As GPT is deployed in settings where informal language is used, interpreting emojis in context will likely become critical. I think that improvement on this emoji riddle task would make GPT better at mimicking human-like communication, as it would be able to understand and respond to various forms of expressions involving emojis. Emojis and their combinations often carry cultural and social meanings. By being adept at emoji riddles, ChatGPT would showcase an understanding of cultural nuances and be more relatable to users.
-
problem-solving: Emoji riddle solving requires i) extracting possible meanings and ii) finding the more suitable association of meaning in the given context (cultural, plateform, etc).
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - Include at least 100 high quality examples (it is okay to only contribute 5-10 meaningful examples and have us test them with GPT-4 before adding all 100)
If there is anything else that makes your eval worth including, please document it below.
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.yaml - Ensure you have the right to use the data you submit via this eval
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
Since we are using Git LFS, we are asking eval submitters to add in as many Eval Samples (at least 5) from their contribution here:
View evals in JSON
{"input":[{"role":"system","content":"You are an emoji riddle solver.
You understand that an emoji riddle consists of finding the word or
group of words associated with an association of emojis that is provided
with the following format: emoji_1 + ... + emoji_n = ? . Your task is to
find the right answer."},{"role":"user","content":"👀 + 🪚 = ? \n Your
answer should strictly only contain the group of words associated with
the answer, no additional words. Don't add `The answer is`. don't add a
period at the end of your answer. everything should be
lowercase"}],"ideal":["seesaw"]}
{"input":[{"role":"system","content":"You are an emoji riddle solver.
You understand that an emoji riddle consists of finding the word or
group of words associated with an association of emojis that is provided
with the following format: emoji_1 + ... + emoji_n = ? . Your task is to
find the right answer."},{"role":"user","content":"❤️ + ✉️ = ? \n Your
answer should strictly only contain the group of words associated with
the answer, no additional words. Don't add `The answer is`. don't add a
period at the end of your answer. everything should be
lowercase"}],"ideal":["love letter"]}
{"input":[{"role":"system","content":"You are an emoji riddle solver.
You understand that an emoji riddle consists of finding the word or
group of words associated with an association of emojis that is provided
with the following format: emoji_1 + ... + emoji_n = ? . Your task is to
find the right answer."},{"role":"user","content":" ⌚️ + 🐶 = ? \n Your
answer should strictly only contain the group of words associated with
the answer, no additional words. Don't add `The answer is`. don't add a
period at the end of your answer. everything should be
lowercase"}],"ideal":["watchdog"]}The Dataset
app: experimental Tauri branch (#50620)
This is experimental support for building the Cody App using Tauri. For an overview of what Tauri is and why I think it will help us with the App, see this Slack message.
To try it out, checkout this branch and then in two separate terminals run:
sg start app
go build \
-o .bin/backend-aarch64-apple-darwin \
-tags dist \
-ldflags '-X github.com/sourcegraph/sourcegraph/internal/conf/deploy.forceType=app' \
./enterprise/cmd/sourcegraph
pnpm tauri dev
This will open a Tauri window connected to your dev server.
We will follow-up to integrate this into sg start app more properly
soon.
./enterprise/dev/app/build-release.sh
This will first invoke esbuild to generate the bundles; then it will run
go build to create the Go backend binary; and then finally it will
invoke pnpm tauri build to produce the macOS app.
Once that command finishes, you'll find the app in
./src-tauri/target/release/bundle/ (make sure you wait for it to
finish, it will open a window and move things around before it is done.)
- Familiarize more folks on the team with this code; add better docs
- Make
sg start appautomatically use Tauri, without needing to e.g. run thepnpm tauri devcommand separately. - Use GitHub actions to start building+releasing versions of this in our CI pipeline
- Make
./enterprise/dev/app/build-release.shwork on Linux - Make
./enterprise/dev/app/build-release.shproduce a Universal macOS binary, not just for Apple Silicon - Start hacking, making improvements to the whole experience :)
- Myself, Juliana, and William are happy with this as a starting point and are able to run/develop with it.
- The changes have limited blast radius, should only affect App and we'll have more time to make improvements before releasing this version to any users.
- We can continue releasing the old-style App version to users just in case we should want/need to create a release before this new version is ready.
Signed-off-by: Stephen Gutekanst [email protected] Co-authored-by: William Bezuidenhout [email protected]
Drow as a playable race (initial commit).
This is the initial commit for a new playable race, the Drow. Drow are also known as dark elves, and share many traits with their surface-dwelling, light-tolerating cousins.
Drow starting stats and abilities almost exactly match that of elves (player and monster), including their material hatred of iron. The differences in this commit that the player will notice, is that Drow elves start with sleep resistance immediately, and will gain poison resistance at expereince level five. They can only be chaotic (or unaligned as an Infidel). Their available roles are Convict, Infidel, (Dark) Knight, Priest/Priestess, Rogue, Ranger, or Wizard. Male and female genders are available.
Drow hate orcs just as their fairer cousins do, but they hate normal elves even more, and elves will grudge drow.
The artifact Grimtooth will warn against drow as well as elves, and can cause drow extra damage from special attacks. The forged artifact Shadowblade is now attuned to the drow race.
Droven mummies and zombies have also been created, since drow can bturned into each. Shopkeepers, priests, and player monsters can also spawn as the droven race. Shopkeeper pricing adjustments match those of elves.
This is just the foundational commit. There's a LOT more work to do here, including drow-specific gear, new material(s), new monsters, and most importantly how light/dark affect drow. Traditionally, drow can tolerate light, even that of the sun if conditioned to it over time. But they prefer the dark, and certain benefits the drow have just aren't available while in the light. Will need to balance how drow operate in other fantasy settings vs how it would equate to EvilHack.
Algebra word problems (#36)
🚨 Please make sure your PR follows these guidelines, failure to follow the guidelines below will result in the PR being closed automatically. Note that even if the criteria are met, that does not guarantee the PR will be merged nor GPT-4 access granted. 🚨
PLEASE READ THIS:
In order for a PR to be merged, it must fail on GPT-4. We are aware that right now, users do not have access, so you will not be able to tell if the eval fails or not. Please run your eval with GPT-3.5-Turbo, but keep in mind as we run the eval, if GPT-4 gets higher than 90% on the eval, we will likely reject since GPT-4 is already capable of completing the task.
We plan to roll out a way for users submitting evals to see the eval performance on GPT-4 soon. Stay tuned! Until then, you will not be able to see the eval performance on GPT-4.
Algebra Word Problems (algebra_word_problems)
This eval contains some algebra word problems that tend to make gpt 3.5 hallucinate, but wouldn't be out of place on a grade school exam. Currently has less than 100 examples, but will add if folks think this is a good eval path to go down.
Poor performance on GPT 3.5 for one, but also would be a great test of increased logical reasoning capabilities of GPT-4 per the release blog post.
Below are some of the criteria we look for in a good eval. In general, we are seeking cases where the model does not do a good job despite being capable of generating a good response (note that there are some things large language models cannot do, so those would not make good evals).
Your eval should be:
- Thematically consistent: The eval should be thematically consistent. We'd like to see a number of prompts all demonstrating some particular failure mode. For example, we can create an eval on cases where the model fails to reason about the physical world.
- Contains failures where a human can do the task, but either GPT-4 or GPT-3.5-Turbo could not.
- Includes good signal around what is the right behavior. This means
either a correct answer for
Basicevals or theFactModel-graded eval, or an exhaustive rubric for evaluating answers for theCriteriaModel-graded eval. - [] Include at least 100 high quality examples
If there is anything else that makes your eval worth including, please document it below.
Insert what makes your eval high quality that was not mentioned above. (Not required)
Your eval should
- Check that your data is in
evals/registry/data/{name} - Check that your yaml is registered at
evals/registry/evals/{name}.jsonl - Ensure you have the right to use the data you submit via this eval
(For now, we will only be approving evals that use one of the existing eval classes. You may still write custom eval classes for your own cases, and we may consider merging them in the future.)
By contributing to Evals, you are agreeing to make your evaluation logic and data under the same MIT license as this repository. You must have adequate rights to upload any data used in an Eval. OpenAI reserves the right to use this data in future service improvements to our product. Contributions to OpenAI Evals will be subject to our usual Usage Policies (https://platform.openai.com/docs/usage-policies).
- I agree that my submission will be made available under an MIT license and complies with OpenAI's usage policies.
If your submission is accepted, we will be granting GPT-4 access to a limited number of contributors. Access will be given to the email address associated with the merged pull request.
- I acknowledge that GPT-4 access will only be granted, if applicable, to the email address used for my merged pull request.
We know that you might be excited to contribute to OpenAI's mission, help improve our models, and gain access to GPT-4. However, due to the requirements mentioned above and high volume of submissions, we will not be able to accept all submissions and thus not grant everyone who opens a PR GPT-4 access. We know this is disappointing, but we hope to set the right expectation before you open this PR.
- I understand that opening a PR, even if it meets the requirements above, does not guarantee the PR will be merged nor GPT-4 access granted.
- I have filled out all required fields in the evals PR form
- (Ignore if not submitting code) I have run
pip install pre-commit; pre-commit installand have verified thatblack,isort, andautoflakeare running when I commit and push
Failure to fill out all required fields will result in the PR being closed.
Since we are using Git LFS, we are asking eval submitters to add in their first 100 JSONL eval lines.
View evals in JSON
{"input": [{"role": "system", "content": "Answer the following question
with a single number and no additional text. You are a helpful
assistant."}, {"role": "user", "content": "If it takes 5 machines 5
minutes to make 5 devices, how long would it take 100 machines to make
100 devices?"}], "ideal": "5"}
{"input": [{"role": "system", "content": "Answer the following question
with a single number and no additional text. You are a helpful
assistant."}, {"role": "user", "content": "What is the sum of 60000,
5000, 400, and 3, with the third value multiplied by 5 before performing
the operation?"}], "ideal": "67003"}
{"input": [{"role": "system", "content": "Answer the following question
with a single number and no additional text. You are a helpful
assistant."}, {"role": "user", "content": "If the sum of the smallest
and largest of three consecutive even numbers is 28, what is the value
of the second largest number in the series?"}], "ideal": "14"}
{"input": [{"role": "system", "content": "Answer the following question
with a single number and no additional text. You are a helpful
assistant."}, {"role": "user", "content": "John is trying to fill a 16
oz. bottle with water. If John fills the bottle at 1 oz per second and
the bottle leaks .2 oz per second, how long would it take for John to
fill the bottle?"}], "ideal": "20"}
{"input": [{"role": "system", "content": "Answer the following question
with a single number and no additional text. You are a helpful
assistant."}, {"role": "user", "content": "Annie is training for a
marathon. She has a weekly training routine, training for five hours a
day on some days and 3 hours a day on the other days. She trains a total
of 27 hours in a seven day week. On how many days does she train for
five hours?"}], "ideal": "3"}
{"input": [{"role": "system", "content": "Answer the following question
with a single number and no additional text. You are a helpful
assistant."}, {"role": "user", "content": "At the start of the year the
ratio of boys to girls in a class is 2 : 1. But now, half a year later,
four boys have left the class and there are two new girls. The ratio of
boys to girls is now 4 : 3. How many students are there altogether
now?"}], "ideal": "28"}Finished Zombie Proficiency & Weakpoint Review (#64194)
- Reviewed all Zombie Weakpoints & Proficiencies
- Renamed Large Humanoids proficiency to Giant Humanoids, to clarify that it does not apply to somewhat large humanoids, like brutes, and only works on hulks and similar.
- Changed description of Natural Armors proficiency, as many enemies that used this proficiency had something more like a thick hide than any kind of shell.
- Renamed Natural Armor weakpoint set (wps_natural_armor) to wps_armored_hide, to better reflect its purpose and to avoid confusion with the unrelated Natural Armor proficiency, as well as to prevent its misapplication to monsters which have more of a carapace or plate armor thing going on. Natural Armors proficiency should be reserved for uniquely resilient armored foes, like kevlar zombies, whereas armored hide applies to anything with a particularly thick hide, even if not outrageously so.
- Edited description of Corrosive Zombie to hint at its thick hide. Corrosive zombie now also trains Natural Armor proficiency.
- Spitter now has big head weakpoint set, based on description.
- All amalgamations now have intro_biology in their families. This should really be on any living creature of flesh and blood, with exceptions only for stuff like robots, physics-defying nether creatures, extra-dimensional anomalies, and the cafeteria meatloaf. I didn't add this to the cocoons because I wasn't sure if that made sense to do.
- Caustic amalgamation now trains biochemistry, like acid zombies do.
- Charged amalgamation now trains electromagnetics, like zapper zombies do.
- Fixed a typo in the description for Zombie Kinderlings.
- Zombie Fiend now trains Ossified Exoskeletons. Thought I added that one earlier.
- Scorched Zombie now gets Armored Hide weakpoints due to its "leathery shell".
- Removed Armored Hides weakpoint set from rust shell zombie and plated zombie. Could possibly apply Ossified Exoskeletons to them, but I'm not sure.
- Slight description tweaks, typo fix.
- Added proficiencies to Aberration and Dissoluted Devourer. Aberration doesn't give zombie bio because it isn't an actual zombie.
- Removed zombie bio from phase skulker, phase shrike, etc, as they aren't actually zombies.
- Gave phase shrike Ossified Exoskeletons proficiency.
- Added basic proficiencies to zombullfrog, frogmother, zombie nemesis, smoker
- Added basic weakpoints to smoker.
- Headless Horror trains giant humanoids proficiency, based on description.
- Removed Malicious Mane's natural armor training and body armor weakpoints, as it had no natural armor (or armor at all, for that matter).
- Added standard proficiencies and weakpoints to all of them.
- Replaced body armor weakpoint set with armored hide.
- Removed military pilot's synthetic armor proficiency
- Gave gastro bufo standard proficiencies and biochemistry.
- Replaced beekeper's body armor weakpoints with armored hide weakpoints
- Added expected proficiencies and weakpoints to pupating hulks, as they were the only pupa zombies that didn't have a copy-from pointing to the base type, and did not include this information. I noticed that most things that disappear on death - boomers, certain cocoons, etc. - tend not to have weakpoints or train proficiencies. Is this an oversight, or is this intentional? For now I left that as is.
- Gave raptors standard and flying proficiencies.
- Electric raptor also teaches electromagnetics, like electric zombies.
-
Removed my Personal Changelog from the Project Directory
-
Fixed Fungal Wretch Typos
-
Linted zed_amalgamations.json
-
MANY Zombie Weakpoint Refinements (& Tests)
- Gave standard weakpoints to standard zombies - manually defined weakpoints for some of the basic zombie models (in zed_misc), like the zombie brute and zombie hulk, is a bit strange, since they have become some of the game's staple enemies. THIS WILL LIKELY EFFECT BALANCE, as these are not only important benchmark enemies, but also copy_from'd by quite a few other enemies. Basic brutes are now somewhat weaker depending on circumstances
- Updated ranged balance test to use enemies with a more uniform form factor, as the high volume of some benchmark enemies lead to counterintuitive results (higher armor enemy taking more damage because it's bigger and easier to shoot). Note that test differences in values aren't all actual "balance changes" but moreso changes to the test itself, so the comparison between old and new isn't 1:1. Test values were only updated on tests that failed for me (I ran the test with 10,000 cycles instead of the usual 200 to be sure the values I got were convergent).
- Added weakpoints and proficiency families to zombies I previously wasn't sure should receive them (mostly ones which self-destruct on death in some way, like boomers). This will make boomers significantly weaker, as they previously had no weakpoints whatsoever.
- Changed boomer stats so no boomer upgrade becomes smaller in volume or lighter in weight than the basic boomer.
- Added an upgrade path for Zombie Miners - they now have a chance to evolve into a shady zombie (most likely), a rust zombie, or just a normal tough zombie, with a ~70% chance not to evolve, on a half-life of 35.
- Rust shell zombies and rust plated zombies get a unique weakpoint category. Similar to bone armor, with the difference that weak points are quite a bit weaker, but the strong point is also a bit stronger.
- Flesh raptors finally have weakpoints, borrowing from the ones used for wasps.
- Removed NOHEAD flag from zombie military pilot, as it very much has a head and there's no reason to believe it to be structurally superfluous, and also fixed them being given erroneous armor weakpoints when they're just in fatigues.
- Lots of other minor weakpoint tweaks/fixes.
-
Revert change to ranged tests that made it run 50 times as long.
-
Update data/json/monsters/zed_amalgamation.json
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
- Update data/json/monsters/zed_children.json
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
-
Update all Range Balance Values
-
Reverted Weakpoint ID Change
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Saloon rework (#1594)
Expands whitesands_surface_camp_saloon to cover a 30x30 footprint and
not be nearly as bad. The previous version had some really glaring
design flaws, like holes in the wall for a bar. On a planet with a
deadly atmosphere. Yeah. Also all the chairs faced the same direction.
You can see how it looks. It's not great.
Here's the new version
Ignore the patches of error, it's purple grass and doesn't display the
icon in sdmm for some reason.
The major changes are: Expanding the building's footprint out to 30x30 Moving the loot behind the building, but locking it behind a shovel of some sort (of which you can go through the ruin to get). Improving the loot a LITTLE
- The map loads although I still haven't managed to get it to load on the proper planet with the spawning verb
The old version was kinda bad. Between the clown and mime masks out front. The small footprint, and the free guns (also out front). This solves those issues kinda while making it bigger.
🆑 add: Camp_Saloon has been expanded, expect frontier luxuries if you find it! /🆑
Co-authored-by: spockye [email protected]
New Ruin: The Beach Town (#1572)
Adds a new beach ruin, the abandoned beachside town
The town is an mostly empty place formerly devoted to tourism and the beloved art of "chilling out". Facets of the life of its inhabitants before their disappearance included drinking, grilling, and swimming off the coast of their fairly large beach. Many interesting things happened on the boardwalk, and a landing pad was present to allow for small ships to dock inside the town.
The loot list is sparse here. I intend for this to mostly be a setpiece for roleplay instead of a loot pinata. There's a good selection of hydroponics seeds and gear, 2 full bar kits, basic kitchen equipment, an autolathe, a few PDAs, a lotta wood, and a jukebox. Also donuts.
- Ruin spawns, nothing is out of whack that shouldn't be.
Continues the trend of making planets more good by adding more content
🆑 add: An oddly empty town has been spotted on beach planets in the area. Check it out spacers. add: Random donut spawners, never eat the same donut two days in a row!
/:cl:
Signed-off-by: Bjarl [email protected]
Ports mothroaches + Moth emotes (#1843)
Can you guess what this PR does? If you answered that it ports this pull request, this pull request, and a partial part of this one too, then you're right!
You can also craft moth plushies now. You just need some cloth, mothroach hide, and a heart!
silly little moth roaches and emotes, who wouldn't want them in the game?
🆑 add: Mothroaches are now a thing add: Moth laughter, chittering and squeaking /🆑
Revert "thermal_core: Do not unload thermal core driver as a module"
thermal_unregister_governors() is not marked with __init annotation anymore and my sorry ass didn't remember during rebase. Revert this broken patch.
This reverts commit e3036b0a6a61076444cf6b4e8dd83e52e581c939.
Signed-off-by: Tashfin Shakeer Rhythm [email protected]
Stupid me #facePalm . KSPe looks for the config file on <add-on-dir>/PluginData , not on <add-on-dir>/Plugins/PluginData! (and I wrote the damned thing…)
For net-lisias-ksp/DistantObject#25
Microing var/static times (~0.015 seconds of init) (#74769)
Moth and I came up with an affront to god and man, and used it to track the time spent creating /static (and in theory /global) variables (this happens right at the start of init) They cost as a sum about 0.05 seconds btw, at least currently.
/datum/timer
var/key
/datum/timer/New(file, line)
src.key = "[file]:[line]"
/datum/timer/proc/operator*(x)
rustg_time_reset(key)
return x
/datum/timer/proc/operator+(x)
var/time = rustg_time_microseconds(key)
world.log << "TIMER: [key]: [time]"
return x
Regex:
var/static/([\w/]+) =
-> var/static/$1 = (new /datum/timer(__FILE__, __LINE__)) * (new /datum/timer(__FILE__, __LINE__)) +
Output on moth's pc looks like this, time in microseconds
Most of this is either icon_states() memes (which appears to be cached btw, that's interesting), or a variation on typecacheof() There is one get_asset_datum call, but that is ALREADY cached and so is just redundant. That's a good 0.01 seconds saved.
The rest of the time here is slightly more interesting.
The majority of typecacheof() is iterating the output of typesof(), a byond internal proc that returns a list of types that either are or are the child of the passed in type. A decent chunk of time here (0.005 seconds, or 10% of the proc) can be saved by unrolling the arguments to the proc. It takes an arbitrary amount of typepaths as input, but we can't like use arglist() here (cause this is an internal "proc"), so instead we try a window of args, passing in null if we start to try and take in too much. Window size matters, zebra fits better into 4 then 5, especially because of how grouping needs to work to make this effect happen. We save about 0.001 for zebra btw, which is around about 7%. It's lower cause we need to group the paths beforehand I think.
The speedup is minor, but it DOES exist. Plus it's fun.
Microing is a hell of a drug
FRM JSON PACK - 20 APR 2023 (V11)
Sorry for the huge delay, was a little bit busy but I'm back with some fixes:
Arcade (fixes): dinore,mslug3lw
Arcade (new): ffightaec2,kf2k2ps2b,kof98eck20,mslug5d,samsh5pf,teot,avengrgsbh,captre,dinoares,dinocunp,dinocx3,dinohced,dinombull,dinopuni,dinox5,fatfurspbs,fatfury3bh,ffightaemgc,jurass99p,kf2k3ps2sp,kof2000otc,kof2001ru,kof2k2plus,kof2kxxx,kof96rss,kof97evn,kof97inv,kof98bc2nd,kof98bc2k2,magdrop3te,sailormnrot,sfa2uhc,sfiii3ws,sfiii4fs,sfiiibh,sfpp,umk3p,xmcotan,xmvsfcph
Genesis (new): sor2fnr,fightvengt,punisor,sonic3kbrc,sor2cc,sor2em,sor2ffc20,sor2tncha,sor2tnwoa,sor2tww,sor2wof1k,tmnttsorp,insanepain,tmntsrr
NES (new): smbtwopla,ducktales2tp,kartfighter,skartfighter,hackmatch,nekkestrbasen,tetristpg,famista93e,tetristpn
SMS (new): alexkidd3f
SMS (fixes): voyage
Arcade (fixes): dinore,mslug3lw
Arcade (new): ffightaec2,kf2k2ps2b,kof98eck20,mslug5d,samsh5pf,teot,avengrgsbh,captre,dinoares,dinocunp,dinocx3,dinohced,dinombull,dinopuni,dinox5,fatfurspbs,fatfury3bh,ffightaemgc,jurass99p,kf2k3ps2sp,kof2000otc,kof2001ru,kof2k2plus,kof2kxxx,kof96rss,kof97evn,kof97inv,kof98bc2nd,kof98bc2k2,magdrop3te,sailormnrot,sfa2uhc,sfiii3ws,sfiii4fs,sfiiibh,sfpp,umk3p,xmcotan,xmvsfcph
Genesis (new): sor2fnr,fightvengt,punisor,sonic3kbrc,sor2cc,sor2em,sor2ffc20,sor2tncha,sor2tnwoa,sor2tww,sor2wof1k,tmnttsorp,insanepain,tmntsrr
NES (new): smbtwopla,ducktales2tp,kartfighter,skartfighter,hackmatch,nekkestrbasen,tetristpg,famista93e,tetristpn
SMS (new): alexkidd3f
SMS (fixes): voyage
SNES (fixes): zenprow,7thsaga,2020bb,acenerae,actrais2j,actrais2u,aerofgt,airdiverj,airdivr2,ajmajonm,andrindy,aokiden,aressh3,astobelx,avspu,ballz3d,barbiesm,barbvac,barkleyu,basload2,bassmc,batblaz,batlcars,batlsoc2,bdodge2,bikedais,bingbing,bluecrys,bof,bofja,brainlrd,brandish,brawlbrou,bretthu,buckrog,bugsbrabu,capcomss,cdalecup,chesterw,contraspd,crayon,ctribe,ctsuba4,daibakjd,daikokaia,daimono2,daimonoga,ddragon5u,deaddanc,dennisu,dmasteru,dokap321a,dolkusay,doraemn3,doraemona,dquest5,dquest12,dreambas,dstall2,ejimu,elfaria,elnard,esbua,estpoli2,exhaust2,f1roc2,fatfury2u,fatfuryu,ffant2a,ffant3a,ffant4ja,ffant5j,ffant6j,ffantmqa,ffantmqj,ffight2u,fghthist,finalstr,firembmn1,fireprw3a,fireprwsa,flashbj,forms95,frontmisa,garou2a,garoua,garousp,gbattle3,gbattle4,genchohi,ggoemkirb,ggoemon2,ggoemon3a,gindamapa,giseiha,gndmxdim,gogoack2,gouketsu,guts,haristad,haristd2,haruaug2,hiryukgf,hiryukhv,homeimpr,itadaki2,jbsuperb,jikkscr2a,jlexct94a,jlexct95,jlprime2,jlssocr,juteisen,karatebu,kawanus2,kidkleet,kingarth,kirbybow,kishinko,koryuki,kotm2j,lastbib3,libertyj,lobo,lockon,lordmonaa,madara2,madoum,majtnseib,mbomber,megamnxua,metalmaru,mickeym2,militia,mjtaika2a,mku,momodhap,monopol2,moritas2,moritash,mother2,mspacmanu,naruhodo,nbaliv96u,nflpro94,niceshot,ninjawaru,pachimo2,pacman2u,paladin,pga96ua,pgaua,picrosv1a,picrosv4a,pinkie,pocky2u,populus2j,powerhir,powyak2b,prinmak,ranmabak,ranmahb2,riserobou,robotrek,rockmans,rockmnx2,rockmnx3,rocko,rokudena,roman3k3,runsaberu,ryukokena,sailormn,sailorsbze,sailorsf,samshou,samspir,sangoku3a,sangoku4a,sanspo,sbm2,sbombmn3j,sdgungx,sf2tua,sf2u,sfamist5,sgenjin,shinmt2b,shodaneka,shotok94,shotoku2,shounin,shushoku,sjinsei2a,sjinsei3a,sjinseia,skeiba2,slamdnk2,slamdunk,slammastu,slayers,sloopzj,smetroidu,smkartu,spleagu3,spleagu4,spleague,spuyopb,srobotex,ssf2u,sshogi2,starocn,street95,super3db,supermjte,superozup,supf1c2,supf1cg,suprinin,suzuka8,tactsocr,taikoris,targa,tecmosbw,teiketsua,tengaim0,tg3kj,tg3ku,tmnttfce,tophant,topman2,ultfight,umizurim,vgundam,votoms,wagyanp,wayneg,winpost2,wizardr6,xak,yokoms2,yokozunaa,ys4
There were some updates to hacks and other stuff (mostly thanks to tobemorecrazy)
gpg-interface: set trust level of missing key to "undefined"
In check_signature(), we initialize the trust_level field to "-1", with the idea that if gpg does not return a trust level at all (if there is no signature, or if the signature is made by an unknown key), we'll use that value. But this has two problems:
-
Since the field is an enum, it's up to the compiler to decide what underlying storage to use, and it only has to fit the values we've declared. So we may not be able to store "-1" at all. And indeed, on my system (linux with gcc), the resulting enum is an unsigned 32-bit value, and -1 becomes 4294967295.
The difference may seem academic (and you even get "-1" if you pass it to printf("%d")), but it means that code like this:
status |= sigc->trust_level < configured_min_trust_level;
does not necessarily behave as expected. This turns out not to be a bug in practice, though, because we keep the "-1" only when gpg did not report a signature from a known key, in which case the line above:
status |= sigc->result != 'G';
would always set status to non-zero anyway. So only a 'G' signature with no parsed trust level would cause a problem, which doesn't seem likely to trigger (outside of unexpected gpg behavior).
-
When using the "%GT" format placeholder, we pass the value to gpg_trust_level_to_str(), which complains that the value is out of range with a BUG(). This behavior was introduced by 803978da49 (gpg-interface: add function for converting trust level to string, 2022-07-11). Before that, we just did a switch() on the enum, and anything that wasn't matched would end up as the empty string.
Curiously, solving this by naively doing:
if (level < 0) return "";
in that function isn't sufficient. Because of (1) above, the compiler can (and does in my case) actually remove that conditional as dead code!
We can solve both by representing this state as an enum value. We could do this by adding a new "unknown" value. But this really seems to match the existing "undefined" level well. GPG describes this as "Not enough information for calculation".
We have tests in t7510 that trigger this case (verifying a signature from a key that we don't have, and then checking various %G placeholders), but they didn't notice the BUG() because we didn't look at %GT for that case! Let's make sure we check all %G placeholders for each case in the formatting tests.
The interesting ones here are "show unknown signature with custom format" and "show lack of signature with custom format", both of which would BUG() before, and now turn %GT into "undefined". Prior to 803978da49 they would have turned it into the empty string, but I think saying "undefined" consistently is a reasonable outcome, and probably makes life easier for anyone parsing the output (and any such parser had to be ready to see "undefined" already).
The other modified tests produce the same output before and after this patch, but now we're consistently checking both %G? and %GT in all of them.
commit: give a hint when a commit message has been abandoned
If we launch an editor for the user to create a commit message, they may put significant work into doing so. Typically we try to check common mistakes that could cause the commit to fail early, so that we die before the user goes to the trouble.
We may still experience some errors afterwards, though; in this case, the user is given no hint that their commit message has been saved. Let's tell them where it is.
Signed-off-by: Jeff King [email protected]
Stops station blueprints from expanding areas of non atmos adjacent turfs. (#74620)
Fixes #74605
the problem starts with detect_room() proc. This proc returns turfs
even those with atmos_adjacent_turfs = null. This means it returns
turfs that has a wall, airlock, window etc i.e. whatever that stops air
from flowing through it. This coupled together with create_area()
causes some wierdness.
Let's take an example
Area A is well defined i.e. it has been created via the station
blueprints and is highlighted in green, Area B however is only
theoretical i.e. we haven't created it yet or we are about to create it.
Now you might be thinking Area A is completely walled & sealed off, it
should be physically impossible to expand it unless we broke down one of
it's walls and so since we are standing in Area B it shoudn't even give
me the option to expand area A Right? right? r.i.g.h.t?
Well PHFUUK. The area editor completely ignores the laws of physics and
allows me expand Area A anyway. This could cause some real power gaming
shit because if you create an area next to an area having an APC you
could use that area power without even making your own apc by simply
expanding that area(like using someone else's wifi from outside their
house without them even knowing)
#73850 accidently built on top of this as it relied on this to detect duplicate APC's but the checks became way too strict as it would check areas of surrounding walls for apc's and throw the conflicting apc error. You can now build room's next to each other even if they have fuctioning apc's however you still can't build rooms in space on top of shuttle walls because that's been the default behaviour for years and hasn't been touched one bit.
🆑 fix: station blueprints no longer expands & detects areas of non atmos adjacent turfs. /🆑
Update README.md (This really bugged me sorry lol)
got rid of the doubling of "GitHub Repository" as the hyperlink text does the job of rendering the text and providing the link. I'm sure it was a typo no biggie, honestly a super trivial edit I'm aware but it was driving me crazy!
from this: If you are interested in learning more about this groundbreaking project, visit their Github repository github repository, where you can find comprehensive information regarding the app's functionalities and technical details. Moreover, you can delve deeper into the training process and database by going through their detailed Technical report, available for download at Technical report.
To this:
If you are interested in learning more about this groundbreaking project, visit their github repository, where you can find comprehensive information regarding the app's functionalities and technical details. Moreover, you can delve deeper into the training process and database by going through their detailed Technical report, available for download at Technical report.
Holy fucking shit fuck this EOW iconcam it's so over ;-;
Fix bugs caused by Preact update
The new Preact version seems to have broken a lot of low-level magic we used. We plausibly shouldn't be using such low-level magic in the first place, but that's a conversation for another day.
In particular:
-
preact.renderseems to replace all ofcontainerNode's contents ifreplaceNodeisn't passed (previously, it would append a child). This is an insane thing to change without any documentation... Maybe I'm misunderstanding it? -
Making a button value an uncontrolled form was a pretty big hack in the first place, but at least it worked. Now that it doesn't, we're giving up and switching to controlled forms, which makes the code a lot nicer, fixes a bug, and I should probably have just done in the first place.