fc1020a8e4
Refactor sequence breaker parsing in getTextGenGenerationData function
2025-05-11 11:07:13 +03:00
2aa5addb1d
Mancer parameters:
...
- Add XTC
- Add DRY
- Remove Mirostat
2025-05-10 19:04:32 -04:00
3be991591f
Remove special handling of nsigma for llama.cpp
...
* 0 now changed to disable/no-op upstream
2025-05-06 14:11:00 -07:00
4a5d0df92f
Translate nsigma 0 to -1 to disable for llama.cpp
2025-05-06 09:31:55 -07:00
bf8b3b5013
Remove tfs_z alias for llama.cpp
...
* This sampler is no longer supported in llama.cpp
2025-05-06 00:39:25 -07:00
bf66a39579
Update llama.cpp textgen settings
...
* Add min_keep, a llama.cpp-exclusive setting for constraining the effect of truncation samplers
* Enable nsigma for llama.cpp, and add llama.cpp alias top_n_sigma, add nsigma to the llama.cpp sampler order block
* Allow a negative value of nsigma as this represents 'disabled' in llama.cpp (while 0 is deterministic)
* Remove tfs and top_a as these are not supported by llama.cpp (tfs was removed, and top_a was never supported)
* Correct the identification string for typical_p in the llama.cpp sampler order block
* Add penalties to the llama.cpp sampler order block
2025-05-06 00:32:29 -07:00
98e96b8c07
Check for error as well when parsing streaming responses
2025-04-30 21:23:13 +08:00
d42a81f97c
New connection manager events, ConnectionManagerRequestService ( #3603 )
2025-03-16 16:58:34 +02:00
7d568dd4e0
Generic generate methods ( #3566 )
...
* sendOpenAIRequest/getTextGenGenerationData methods are improved, now it can use custom API, instead of active ones
* Added missing model param
* Removed unnecessary variable
* active_oai_settings -> settings
* settings -> textgenerationwebui_settings
* Better presetToSettings names, simpler settings name in getTextGenGenerationData,
* Removed unused jailbreak_system
* Reverted most core changes, new custom-request.js file
* Forced stream to false, removed duplicate method, exported settingsToUpdate
* Rewrite typedefs to define props one by one
* Added extractData param for simplicity
* Fixed typehints
* Fixed typehints (again)
---------
Co-authored-by: Cohee <18619528+Cohee1207@users.noreply.github.com >
2025-03-03 10:30:20 +02:00
5ecfbf3933
Fix element id for nsigma
2025-03-02 23:01:34 +02:00
810d954d12
Add missing samplers to TC restore preset
2025-03-02 00:26:47 +01:00
5f024823a9
Improve JSdocs
2025-02-25 10:22:03 +00:00
092ef26144
Exported getTextGenServer, extractMessageFromData, getPresetManager methods. Added optional parameters to some methods for generic usage.
2025-02-25 13:20:06 +03:00
890d10d811
Work on tl
2025-02-16 18:43:57 +07:00
1f41124844
Remove pointless return
2025-02-16 02:13:08 +02:00
23f0b6ed09
Add a blank line
2025-02-15 12:38:45 +02:00
c87e203b4a
Refactor banned strings kill switch functions into a single toggle function
2025-02-15 12:37:48 +02:00
c37c9051a6
Fix bug when only global bans are defined
2025-02-15 00:34:07 +02:00
8a4cf86b65
Integrate killswitch into the block
2025-02-15 00:28:38 +02:00
a2a15b9978
Merge branch 'Banned-Strings-Killswitch' into tc-global-banlist
2025-02-15 00:12:16 +02:00
412d638e9e
Text Completion: Add global banned strings list
2025-02-13 22:07:36 +02:00
d5959a4681
Add awaits to event emissions
2025-02-08 19:13:05 +02:00
c5dad20fc4
add killswitch for banned strings
2025-02-02 12:00:01 +09:00
753a99faf9
Support reasoning for OpenRouter text completion
2025-01-31 21:41:56 +09:00
e3f0a8d35b
TC: Trim spaces in server URLs
2025-01-30 00:25:02 +02:00
afae8d02be
The THONKening
2025-01-23 02:52:52 +02:00
e4290140bc
TC: Remove -1 seeds from request body
...
Closes #3336
2025-01-22 22:45:30 +02:00
d14f2f3c77
llama.cpp: fixed wording in comment for logprobs
2024-12-27 01:15:35 -05:00
77414045d9
llama.cpp: fixed logprobs for newest server version
2024-12-27 01:00:55 -05:00
4232f6c5f4
More null checks for llamacpp logprobs parser
2024-12-20 22:00:36 +02:00
6193eb14da
Add DRY to llama.cpp sampler order
2024-12-18 19:04:49 -08:00
cd0b834291
Implement "except" mode for type-specific controls in settings
2024-12-13 20:58:16 +02:00
3167019faf
Add generic text completion API type (100% OAI compatible)
2024-12-13 01:12:10 +02:00
cc73a45d1f
Make lint happy
2024-12-09 17:32:23 +02:00
bcbfcb87b5
aphrodite: send an empty sampler priority list if using the default order
2024-12-09 14:05:33 +00:00
323b9407df
Small cosmetic fixes
2024-12-09 15:26:57 +02:00
3ed5d892f7
Fix bugs with the settings
2024-12-09 15:13:19 +02:00
add108b821
fix the JS issue where both ooba and aphro were using the same container ID
2024-12-07 12:36:10 +00:00
a1115ab4ff
Merge remote-tracking branch 'alpin/staging' into sampler_order
2024-12-07 12:27:55 +00:00
77841dbc21
Add types for SillyTavern.getContext
2024-12-06 16:41:26 +02:00
2ef095c273
Merge branch 'release' into staging
2024-12-04 01:35:50 +02:00
e6be28acea
llama.cpp: Don't send empty DRY sequence breakers
...
Fixes #3048
2024-12-04 01:32:27 +02:00
7ce2841588
Add trim_stop to koboldcpp requests
...
Fixes #3143
2024-12-03 14:49:20 +00:00
80c8e83f09
use strings instead of IDs
2024-12-03 01:46:51 +00:00
e664806990
add DRY, skew, and no repeat ngram penalty
2024-11-25 22:28:58 +00:00
fdd798e2da
Merge branch 'staging' into top-nsigma
2024-11-25 00:08:38 +02:00
0383ea52e9
Linting and commenting
...
Linting and commenting
Linting and commenting
Linting and commenting
Linting and commenting
2024-11-22 11:55:27 +11:00
ec4cff72fe
feat: add top nsigma sampler support
2024-11-19 23:05:23 +00:00
c599f46269
Remove best_of for VLLM
...
It defaults to `n` anyways, and for some reason it was being incorrectly
received on 0.6.4.post1
2024-11-16 02:05:35 -08:00
07feccbe35
koboldcpp: parse logprobs
2024-11-01 11:38:31 +02:00
