67e28d2b5c
Typical sampling needs to use nansum instead of sum
...
If `probs` is zero then `log_probs` will be negative infinity, and the
calculation of `neg_entropy` would then give NaN because zero times
infinity is a mathematically indeterminate value.
We need to use nansum so that those NaN values are treated as zeros to
ignore them in the entropy calculation.
2022-03-28 00:02:31 -04:00
77ae893f4d
Merge pull request #106 from VE-FORBRYDERNE/typical
...
Typical sampling
2022-03-28 00:14:09 +02:00
e2cd49d552
Typo fix in `TypicalLogitsWarper`
2022-03-27 17:08:57 -04:00
bbd0a83fef
Fix `TypicalLogitsWarper` argument typing
2022-03-27 16:59:23 -04:00
d5989d4c62
Hide division by zero warning in JAX typical filter
...
This warning happens when `np.log` gets an input containing zeros.
In that case, NumPy will throw a warning and output negative infinity.
Negative infinity is the correct behaviour here, so we can safely ignore
the warning.
2022-03-27 16:57:12 -04:00
20e48b11d7
Typical sampling
2022-03-27 16:25:50 -04:00
e4c72ca2e5
Merge pull request #104 from VE-FORBRYDERNE/retry-randomgame
...
Allow regenerating random story using Retry button
2022-03-24 12:57:04 +01:00
0348970b19
Make sure AI is not busy when using retry to regenerate random story
2022-03-23 22:09:35 -04:00
4832dd6f37
Allow regenerating random story using Retry button
...
Commit b55e5a8e0b
2022-03-23 13:39:46 -04:00
38d78d10db
Merge pull request #103 from VE-FORBRYDERNE/neox
...
Divide GPT-NeoX replicated bias layers by 4 again instead of by 8
2022-03-21 02:19:32 +01:00
cf99f02ca5
Merge branch 'main' into united
2022-03-20 19:22:53 +01:00
20eab085dd
Fix AutoSave Toggle
2022-03-20 19:12:11 +01:00
5c795609e4
KML Fix
2022-03-20 13:10:56 +01:00
73aecc0510
Divide NeoX replicated bias layers by 4 again instead of by 8
2022-03-20 01:04:55 -04:00
f1487a4551
New Linux Runtime
2022-03-20 00:00:21 +01:00
a7f652f293
Merge pull request #101 from VE-FORBRYDERNE/neox
...
GPT-NeoX-20B support in Colab TPU instances
2022-03-19 09:56:15 +01:00
05fc46b253
Changing this again to divide by 8
2022-03-19 02:09:41 -04:00
b1125a6705
Add EOS and padding token to default NeoX badwords
2022-03-19 01:30:02 -04:00
6c20d0d657
Nevermind, dividing by 4 is actually correct...
2022-03-19 00:55:04 -04:00
f16b61ec77
Should divide NeoX replicated parameters by 8 (not by 4)
...
Also, suppresses the PyTorch 1.11 warning about transposing tensors with
ndim != 2 in the new code
2022-03-19 00:48:33 -04:00
c2c139e940
Change default PE type for NeoX to `neox_rotary`
2022-03-19 00:26:04 -04:00
85a4959efa
Merge branch 'united' into neox
2022-03-18 11:19:03 -04:00
f581fe89cb
Torch version changes
2022-03-17 21:11:36 +01:00
9e9c1c3fe0
Merge pull request #100 from VE-FORBRYDERNE/patch
...
Add PyTorch 1.11 support for lazy loader
2022-03-17 21:06:38 +01:00
c444260eac
Silence PyTorch warning about transposing tensors with dimension != 2
2022-03-17 15:16:56 -04:00
ef21ab9c91
PyTorch 1.9 lazy loader compatibility bugfix
2022-03-17 14:10:51 -04:00
eaf190469d
Add PyTorch 1.11 support for lazy loader
2022-03-17 12:51:41 -04:00
9235754eb9
Dependency Fixes
2022-03-17 00:35:59 +01:00
a3e5e052b3
Newer umamba + slope tweak
2022-03-16 18:34:02 +01:00
95c4251db9
Print two newlines before loading HF models
2022-03-15 13:58:53 -04:00
9e2848e48f
Show parameter count when loading GPT-NeoX in Colab TPU instance
2022-03-15 13:55:27 -04:00
9dc48b15f0
Add custom badwords and pad token ID for GPT-NeoX
2022-03-14 23:31:49 -04:00
88f247d535
GPT-NeoX-20B support in Colab TPU instances
2022-03-14 23:14:20 -04:00
4892556059
Model saving for colab mode
2022-03-13 11:22:44 +01:00
ccadeabbde
Merge pull request #99 from VE-FORBRYDERNE/model-patch
...
Model loading fixes
2022-03-13 11:10:15 +01:00
2b8c46338e
Change current working directory to KoboldAI folder
2022-03-13 01:22:11 -05:00
48d07adb54
Also fallback to generic GPT2 tokenizer in Colab TPU instances
2022-03-12 23:19:35 -05:00
d29a629320
Merge pull request #98 from ebolam/united
...
Fix for retry
2022-03-12 16:52:07 +01:00
45eed78d21
Merge branch 'united' of https://github.com/ebolam/KoboldAI into united
2022-03-12 10:33:01 -05:00
b55e5a8e0b
Retry Bug Fix
2022-03-12 10:32:27 -05:00
2e1b3c82f9
Merge pull request #97 from ebolam/united
...
Fix for retry causing issues for future redo actions
2022-03-11 17:41:49 +01:00
ae854bab3d
Fix for retry causing issues for future redo actions
2022-03-11 11:40:55 -05:00
2c66461c14
Merge pull request #96 from VE-FORBRYDERNE/dlpack
...
Use DLPack to convert PyTorch tensors to JAX arrays
2022-03-10 22:00:38 +01:00
a99eb8724d
Use DLPack to convert PyTorch tensors to JAX arrays
2022-03-10 15:12:42 -05:00
b02d5e8696
Allows missing model_config again
2022-03-10 19:59:10 +01:00
172a548fa1
Fallback to generic GPT2 Tokenizer
2022-03-10 19:52:15 +01:00
68281184bf
Remove Lowmem from TPU
2022-03-09 19:21:15 +01:00
9dee9b5c6d
Ignore incorrect problems
2022-03-09 12:03:37 +01:00
a28e553412
Remove unused gettokenids
2022-03-09 11:59:33 +01:00
7434c9221b
Expand OAI Setting Compatibility
2022-03-07 08:56:47 +01:00
Gnome Ann
henk717
ebolam