John Smith
32904da1ff
fix bug on triton matmul
2023-04-07 15:50:55 +08:00
John Smith
dba3773b30
add triton backend support for v2 model
2023-04-07 15:34:06 +08:00
John Smith
9351f49542
merge pull request in new branch
2023-04-07 10:40:24 +08:00
John Smith
85e9cf004a
Merge pull request #59 from yamashi/main
...
Add flash attention
2023-04-07 10:05:18 +08:00
yamashi
c5aa7fb695
Update finetune.py
2023-04-07 00:43:36 +02:00
yamashi
95cd390d25
Update Finetune4bConfig.py
2023-04-07 00:43:15 +02:00
yamashi
778035152d
Update arg_parser.py
2023-04-07 00:42:34 +02:00
yamashi
2bf5d42f28
Add position_ids to flash attention
2023-04-06 17:46:15 +02:00
yamashi
7770e76c9c
Fix args of flash attention
2023-04-06 17:32:01 +02:00
yamashi
30bf938d03
Update requirements.txt
2023-04-06 13:50:25 +02:00
yamashi
7b18b39dd8
Create llama_flash_attn_monkey_patch.py
2023-04-06 13:49:36 +02:00
yamashi
3ea18575c7
Use flash attention monkeypatch
2023-04-06 13:49:12 +02:00
John Smith
8020b3ec3b
Update README.md
2023-04-06 13:57:32 +08:00
John Smith
9a02a88fb8
add patch for encode function to remove eos token at the beginning of left side
2023-04-06 12:56:27 +08:00
John Smith
085d9556f9
fix bug
2023-04-06 10:46:42 +08:00
Andrey Glushenkov
c991e2a091
Delete triton_test.py
2023-04-06 02:39:40 +03:00
Andrey Glushenkov
0d271d5d90
Add files via upload
...
Fix triton kernels
2023-04-06 02:38:06 +03:00
Andrey Glushenkov
4a2d23aa29
Delete autograd_4bit.py
...
File moved to autograd_4bit module
2023-04-06 02:31:06 +03:00
Andrey Glushenkov
f20570343f
GPTQv2 support
...
GPTQv2 support.
1. Adds dependency on `triton`
2. Refactors autograd_4bit to include both GPTQv1 and GPTQv2
3. Introduces new environment variable GPTQ_VERSION to select autograd_4bit version
4. Fixes triton kernels
5. Matrix multiplications are in fp16
2023-04-06 02:29:36 +03:00
John Smith
86387a0a35
update multi gpu support in finetune.py
2023-04-03 23:55:58 +08:00
John Smith
5655f218ed
add g_idx buffer.\nadd triton matmul utils for future support.
2023-04-02 21:29:06 +08:00
John Smith
f3a25342e1
fix device_map bug when using lora_apply_dir
2023-03-31 19:44:36 +08:00
John Smith
00bf0a1e1b
Update README.md
2023-03-31 14:17:35 +08:00
John Smith
dd0efc721f
Merge pull request #47 from winglian/better-gpt4all
...
fix gpt4all training to more closely match the released logic, other small fixes and optimizations
2023-03-31 11:20:03 +08:00
Wing Lian
8791eaee9a
fix gpt4all training to more closely match the released logic, other small fixes and optimizations
2023-03-30 22:40:40 -04:00
John Smith
878eada8dd
add amp_wrapper for autocast support.
2023-03-30 19:57:19 +08:00
John Smith
b3c91a5af5
Merge pull request #45 from winglian/fix-missing-bracket
...
fix missing paren
2023-03-30 13:53:55 +08:00
Wing Lian
e744aec8bf
fix missing paren
2023-03-29 23:40:30 -04:00
John Smith
8db4633d84
Update README.md
2023-03-30 11:24:25 +08:00
John Smith
8a62560e6c
add offload support
2023-03-30 11:21:21 +08:00
John Smith
32976f91c4
Merge pull request #42 from winglian/multigpu-fix
...
better multi-gpu support, support gpt4all training data
2023-03-30 00:03:27 +08:00
Wing Lian
b7361da58a
better multi-gpu support, support gpt4all training data
2023-03-29 11:21:47 -04:00
John Smith
0fdae9224c
optimized groupsize backward for performance
2023-03-29 17:44:51 +08:00
John Smith
5986649b37
Update README.md
2023-03-29 14:46:28 +08:00
John Smith
1c02d4262d
add resume checkpoint to continue a training
2023-03-29 14:35:39 +08:00
John Smith
2a1cb42966
add padding support as an option
2023-03-29 11:20:16 +08:00
John Smith
cff57ebfa4
Merge pull request #39 from winglian/fix-prompt-eos-token
...
properly include the eos token so inference doesn't blabber on
2023-03-29 10:35:46 +08:00
Wing Lian
daad59f8ef
properly include the eos token so inference doesn't blabber on
2023-03-28 20:53:16 -04:00
John Smith
1719bd0ce3
fix bug
2023-03-29 08:09:40 +08:00
John Smith
1043ded7d9
Merge branch 'main' of github.com:johnsmith0031/alpaca_lora_4bit
2023-03-29 01:26:20 +08:00
John Smith
d28ee06202
fix bug
2023-03-29 01:25:37 +08:00
John Smith
b5e3dae573
Merge pull request #34 from winglian/v2-fixes
...
fixes for most recent update
2023-03-28 23:49:56 +08:00
Wing Lian
b47da33084
fixes for most recent update
2023-03-28 10:56:35 -04:00
John Smith
234004ceb5
fix bug
2023-03-28 22:05:18 +08:00
John Smith
f26615fc0c
fix bug
2023-03-28 21:47:22 +08:00
John Smith
0768d0fdff
update finetune data format
2023-03-28 21:45:33 +08:00
John Smith
8a6c8661df
Merge branch 'main' of github.com:johnsmith0031/alpaca_lora_4bit
...
merged
2023-03-28 21:14:35 +08:00
John Smith
211af574b6
fix bug
2023-03-28 21:12:51 +08:00
John Smith
ac07457473
Update README.md
2023-03-28 20:44:02 +08:00
John Smith
bff039de95
add v2 model support
2023-03-28 20:33:55 +08:00
