henk717
e799178436
Merge pull request #423 from one-some/fix-introduced-bugs
...
Fix streaming bugs
2023-07-25 18:58:16 +02:00
henk717
4497b09056
Merge pull request #424 from one-some/qptq-lazy
...
gptq is soooo lazy
2023-07-25 02:59:50 +02:00
somebody
c80de5120c
Cleanup
2023-07-24 19:45:33 -05:00
somebody
ad4528b5a6
critical change
2023-07-24 17:17:57 -05:00
somebody
a73420c49c
really really really sketchy breakmodel implementation
...
im gonna go lie down for an extended period of time
2023-07-24 17:15:59 -05:00
somebody
ec040620ec
Merge branch 'united' of https://github.com/henk717/KoboldAI into qptq-lazy
2023-07-24 13:26:16 -05:00
somebody
34aa333c44
Last debug
2023-07-24 13:11:06 -05:00
somebody
43a4abaf63
Remove even more debug
2023-07-24 13:10:33 -05:00
somebody
929917efe9
Remove shrieking
2023-07-24 13:09:43 -05:00
somebody
4a6cccb002
Import fix
2023-07-24 13:09:15 -05:00
somebody
a6aafb2525
GPTQ: Patch QuantLinear to not use CPU RAM
2023-07-24 13:07:30 -05:00
somebody
9cc6972c1c
Shh!
2023-07-24 11:30:33 -05:00
somebody
30640acca7
Editor: Don't allow editing or syncing during generation
...
Only bad things can come from that!
Also filter out stream buffer when fixing dirty game text (just in
case!)
2023-07-24 11:26:20 -05:00
somebody
fc7fa991d5
Streaming: Fix streaming not being cleaned up before commentator speaks
...
super duper critical
2023-07-24 10:57:24 -05:00
henk717
04a5e05692
Merge pull request #422 from one-some/fix-prioritization
...
Fix prioritization (probably)
2023-07-24 17:40:36 +02:00
somebody
2fb877db40
Backends: Probably fix sorting
2023-07-24 10:28:22 -05:00
somebody
81e4c8a807
Backends: Fix GPTQ priority
2023-07-24 10:25:44 -05:00
somebody
1df03d9a27
Basic
2023-07-23 20:54:04 -05:00
Henk
30495cf8d8
Fix GPT2
2023-07-24 02:05:07 +02:00
Henk
9fc9cb92f7
Fancy streaming by default
2023-07-24 01:39:16 +02:00
henk717
c60a29e544
Merge pull request #419 from one-some/streaming-fix-2
...
Rework single-gen streaming
2023-07-24 01:31:51 +02:00
Henk
70d2da55e5
Readme changes
2023-07-24 01:03:46 +02:00
somebody
f4593ed04b
Streaming: Fix bad streamingwindow sync
...
sometimes the good solution is the best solution
2023-07-23 17:50:53 -05:00
somebody
8de610df8c
Streaming: Rework single-gen streaming
...
Now has its own packet and element seperate from actions. Fixes several
bugs related to desyncing. Also adds smooth typing effect
(fixes https://github.com/henk717/KoboldAI/issues/263 )
2023-07-23 17:32:52 -05:00
Henk
a963c97acb
Make 4-bit the default part 2
2023-07-24 00:06:20 +02:00
Henk
3409853dfc
Remove GPTQ for Colab
2023-07-23 23:26:35 +02:00
Henk
70dddf9fdc
Prioritize GPTQ
2023-07-23 23:22:02 +02:00
henk717
ae9ec38ae2
Merge pull request #418 from one-some/united
...
index prioritization
2023-07-23 23:09:42 +02:00
Henk
0f913275a9
4-bit as Default
2023-07-23 23:08:11 +02:00
somebody
3aa677ce11
Indexed prioritization
...
not a great way to get that data
2023-07-23 16:04:23 -05:00
Henk
89637ae9d7
GPTQ Requirements
2023-07-23 22:51:47 +02:00
henk717
1facc73b66
Merge pull request #367 from 0cc4m/4bit-plugin
...
GPTQ module
2023-07-23 22:32:20 +02:00
0cc4m
73953068c0
Remove exllama backend, pending further fixes
2023-07-23 22:12:31 +02:00
0cc4m
973aea12ea
Only import big python modules for GPTQ once they get used
2023-07-23 22:07:34 +02:00
0cc4m
49740aa5ab
Fix ntk alpha
2023-07-23 21:56:48 +02:00
Henk
e33a58b74a
Adventure stoppers = regex
2023-07-23 20:45:48 +02:00
henk717
d70481874c
Merge pull request #334 from YellowRoseCx/YellowRoseCx-advpatch-1
...
Add AdventureMode stopper
2023-07-23 20:38:51 +02:00
henk717
66fb8b8937
Merge pull request #415 from one-some/fixes-forever
...
Fixes forever
2023-07-23 20:05:38 +02:00
henk717
b3b67bf50d
Merge pull request #417 from LostRuins/concedo_united
...
updated lite to v50
2023-07-23 16:11:23 +02:00
Concedo
dd8e5f5d05
updated lite to v50
2023-07-23 21:40:08 +08:00
0cc4m
31a984aa3d
Automatically install exllama module
2023-07-23 07:33:51 +02:00
0cc4m
a9aa04fd1b
Merge remote-tracking branch 'upstream/united' into 4bit-plugin
2023-07-23 07:18:58 +02:00
0cc4m
09bb1021dd
Fallback to transformers if hf_bleeding_edge not available
2023-07-23 07:16:52 +02:00
0cc4m
748e5ef318
Add sliders for exllama context size and related methods
2023-07-23 07:11:28 +02:00
Henk
7a5d813b92
Reimplement HF workaround only for llama
2023-07-22 16:59:49 +02:00
Henk
8dd7b93a6c
HF's workaround breaks stuff
2023-07-22 16:29:55 +02:00
Henk
fa9d17b3d3
HF 4.31
2023-07-22 15:25:14 +02:00
somebody
bc8ba91429
Private Mode improvements
...
- cleaned up ui a bit
- disallow setting empty password (user could press the hotkey by
mistake and be confused), giving the user a notification if they try
to.
- gives the login attempter a message if the password is incorrect, as
well as screaming in the logs
- obscures story name in page title
2023-07-21 21:44:10 -05:00
Henk
7823da564e
Link to Lite
2023-07-22 04:04:17 +02:00
somebody
a93c9d20b1
Don't let logo container gobble up clicks
2023-07-21 19:01:51 -05:00