hiyouga
|
f8497921fe
|
fix #4242
Former-commit-id: 577de2fa078ee62d61bca24b75fa59590e2ef66b
|
2024-06-12 16:50:11 +08:00 |
|
Arthur Kim
|
bebca6d01c
|
Support vllm==0.5.0
Former-commit-id: d65a3f7cb6dcd34e9a3ff9e30dffc52a7671d50b
|
2024-06-12 16:49:12 +09:00 |
|
ancv
|
045eb155a2
|
implement efficient packing without cross-contamination attention
Former-commit-id: b2c367bc61c2778dc359613dca496d9e134c2743
|
2024-06-12 11:56:01 +07:00 |
|
hoshi-hiyouga
|
fecb9c9a76
|
Merge pull request #4204 from dignfei/main
fixbug：llama3在增量预训练时应该使用<|end_of_text|>标识文本的结束
Former-commit-id: 9049aab911fa3fbe9cd463ae3805f81b484da052
|
2024-06-11 17:06:10 +08:00 |
|
hoshi-hiyouga
|
bf3de9bfe8
|
Update pretrain.py
Former-commit-id: 0c292332374fb96c3fc753abde42d070a0c1dca8
|
2024-06-11 17:02:14 +08:00 |
|
hiyouga
|
8c574eb3cb
|
fix deepspeed version
Former-commit-id: cca6f351081903ca3b5f79f10accc1bbbae0ee61
|
2024-06-11 16:52:36 +08:00 |
|
d
|
da39715085
|
经过大量的增量预训练，进行对比试验，发现这个bug：llama3在预训练时使用的tokenizer.eos_toke是'<|end_of_text|>' ，这里在每条数据后面也得用这个，而不是'<|eot_id|>'，否则很容易导致严重的性能下降
Former-commit-id: 6979f3f8480755604d8aea8164f6418126e094c5
|
2024-06-11 16:23:40 +08:00 |
|
hiyouga
|
5834651c4a
|
fix #4198
Former-commit-id: 89f2bd8c8c035181927bd530a7ffc733407d674c
|
2024-06-11 15:38:38 +08:00 |
|
hiyouga
|
53de7f7cc3
|
tiny fix
Former-commit-id: 90e14a960d1437a16d35dbabbb8aa50714583d3a
|
2024-06-11 12:48:53 +08:00 |
|
hiyouga
|
e3baa5aa08
|
tiny fix
Former-commit-id: 3f24337a8a995b145b1e8075bc23878eaa363844
|
2024-06-11 01:04:16 +08:00 |
|
hiyouga
|
d6632fefc9
|
set dev version
Former-commit-id: 91e62a098fd997d0d1d12baef64d089aabc01fba
|
2024-06-11 00:50:53 +08:00 |
|
hiyouga
|
75e1bbf128
|
release v0.8.1
Former-commit-id: 2b6ebd6b51133cf114d6f0e8605ad2bb26aa6d65
|
2024-06-11 00:44:26 +08:00 |
|
hiyouga
|
2f164c2c41
|
fix #4160
The split heads should be concatenated in dim=2
Former-commit-id: a793e8456b664ea0b48f0ba162999f18d06b4c2f
|
2024-06-11 00:37:17 +08:00 |
|
hiyouga
|
144544cd37
|
update evaluator
Former-commit-id: 0012762b04d4082a3a7d98889cc3024adc92fc44
|
2024-06-10 23:56:00 +08:00 |
|
hiyouga
|
3b244a69dc
|
fix #2666
Former-commit-id: c907d816670975daa900898660d3503708b7fc37
|
2024-06-10 21:24:15 +08:00 |
|
mMrBun
|
b6d63b3324
|
Optimize the handling of QWEN2 in scenarios involving multiple tool calls.
Former-commit-id: 950e360ca00c29febadc14d5995de7d57b5c43a7
|
2024-06-10 02:00:14 +08:00 |
|
mMrBun
|
3f11ab800f
|
Removed unnecessary comments.
Former-commit-id: 6ed0b0c800d416379acf8395aa852c188c107eb9
|
2024-06-09 18:25:22 +08:00 |
|
mMrBun
|
daf472994d
|
Merge branch 'hiyouga:main' into main
Former-commit-id: 0f2609ce19492f0bab9b4880ded228b5513e5907
|
2024-06-09 18:17:24 +08:00 |
|
mMrBun
|
18a86ea104
|
Implemented the tool_formatter and tool_extractor for glm4 tool_format
Former-commit-id: cb1cbcb293917e960cad8f0eac7a11a122ab644a
|
2024-06-09 18:16:15 +08:00 |
|
hiyouga
|
1a261add61
|
fix llamafactory-cli env
Former-commit-id: 972ec9c668de1a9b6d872187dbc0c1d94f6fec6b
|
2024-06-08 07:15:45 +08:00 |
|
hiyouga
|
de3400a521
|
set dev version
Former-commit-id: 3ac11e77cccf686e0da499bd152997133b49a265
|
2024-06-08 06:46:09 +08:00 |
|
hiyouga
|
ce40d12692
|
release v0.8.0
Former-commit-id: 5aa4ce47567146cd97c61623018153b41d7c1278
|
2024-06-08 05:20:54 +08:00 |
|
hiyouga
|
4f0ce9be4e
|
reorganize adapter code
Former-commit-id: 54cd743ebfbd296ae9eaf10c33f59e127f451785
|
2024-06-08 00:47:23 +08:00 |
|
hoshi-hiyouga
|
bad35d1730
|
fix #4139
Former-commit-id: cfd62283a9772fc854b852d2a1b71699f79a0048
|
2024-06-08 00:45:02 +08:00 |
|
hiyouga
|
a8318723a4
|
add resume args in webui
Former-commit-id: 06e5d136a4916413d1c116e341ba7d5136d7748a
|
2024-06-08 00:22:16 +08:00 |
|
hiyouga
|
d79222894c
|
fix #4137
Former-commit-id: 8bf9da659cc789dbdd909930f79ac59539b51127
|
2024-06-07 19:16:06 +08:00 |
|
hiyouga
|
ca9468ff04
|
tiny fix
Former-commit-id: f8d8690bf4c2981f3151b4ccf07daeb4f3cd38a9
|
2024-06-07 05:19:21 +08:00 |
|
hiyouga
|
4f3c89a6eb
|
fix ppo trainer save zero3 model
accelerator.get_state_dict(ds_model) should be called at all ranks
Former-commit-id: 4489d73ac75c8dbc002fc16c854148994d432c3a
|
2024-06-07 05:14:19 +08:00 |
|
hiyouga
|
f76d427332
|
fix ppo in trl 0.8.6
Former-commit-id: 2702d7e952523b584d67c8901888b492d4a79b14
|
2024-06-07 04:48:29 +08:00 |
|
hiyouga
|
d3196318be
|
fix #4120
Former-commit-id: f9e818d79cf686cb34789327add7ed1f749966c6
|
2024-06-07 04:18:05 +08:00 |
|
hiyouga
|
c6f5f69644
|
update data processors
Former-commit-id: ccc8b64cc21417cfd88d9b5e9bd62e15c1e834a1
|
2024-06-07 04:15:40 +08:00 |
|
hoshi-hiyouga
|
4953ded639
|
Merge pull request #4009 from AlongWY/main
supervised packing with greedy knapsack algorithm
Former-commit-id: 181dbb0d05655c6b46a3859cb47cfa3f86ef8abd
|
2024-06-07 03:48:46 +08:00 |
|
hoshi-hiyouga
|
e3ef239bc0
|
Update supervised.py
Former-commit-id: c09ad8bab38bc2f151da3a924eba225111af2481
|
2024-06-07 03:42:08 +08:00 |
|
hoshi-hiyouga
|
fd7bd911a6
|
Update supervised.py
Former-commit-id: 788e8232fc4ed58ab2439a9bc2e38f64e12c6eb3
|
2024-06-07 03:38:23 +08:00 |
|
hoshi-hiyouga
|
21df5f0bd0
|
Update supervised.py
Former-commit-id: 8cecade7082a52f413517ea20b1c5dd812db8e53
|
2024-06-07 03:38:04 +08:00 |
|
hiyouga
|
8a0263551d
|
add qwen2 models
Former-commit-id: 8e95648850fdd5075724359ffdb22beb48b75952
|
2024-06-07 00:22:57 +08:00 |
|
hiyouga
|
8da149ba40
|
rename files
Former-commit-id: 74f96efef9bcd63f65d0190c901ff9be54ccd350
|
2024-06-07 00:09:06 +08:00 |
|
hiyouga
|
fd2c64315b
|
add DISABLE_TORCHRUN option
Former-commit-id: 45d8be8f93188e6e9d1e0841c2736dce0c1ef7d2
|
2024-06-06 23:44:58 +08:00 |
|
hoshi-hiyouga
|
77b70664e5
|
Merge pull request #4082 from MengqingCao/bugfix
Fix #4077
Former-commit-id: 55c18c49b00a9c13ad42c1b09cd5844253876267
|
2024-06-06 23:38:40 +08:00 |
|
hoshi-hiyouga
|
9bebdeabda
|
Update cli.py
Former-commit-id: 751dd77bc076af3a13884a609ec71c089c516f1f
|
2024-06-06 23:38:09 +08:00 |
|
hiyouga
|
368695483d
|
fix ppo+zero3 #3108
Former-commit-id: 76c61905b20f69fac5c7a6c4ea9450bf33d3b1f2
|
2024-06-06 23:30:07 +08:00 |
|
hiyouga
|
6cbc66a602
|
fix torch gc
Former-commit-id: 451b6693c0cb86cc9ac03d1a9389cf1fd2b918ec
|
2024-06-06 20:30:25 +08:00 |
|
hiyouga
|
e0aadd4b34
|
fix ppo dataset bug #4012
Former-commit-id: 149610c636bbb974e546d13fa302884ea65a6d38
|
2024-06-06 19:03:20 +08:00 |
|
hiyouga
|
e898d8bbc4
|
update trainers
Former-commit-id: fad2591e314093335ef1c301d0a70f0cbe935728
|
2024-06-06 18:45:49 +08:00 |
|
hiyouga
|
f2edacb02d
|
fix base64 image read #4061
Former-commit-id: 67aa78cde0d89e9f4b5640bce5fc506ef73ff7b6
|
2024-06-06 17:29:19 +08:00 |
|
hiyouga
|
cceff9f520
|
lora modules: all by default
Former-commit-id: cae47379079ff811aa385c297481a27020a8da6b
|
2024-06-06 03:53:28 +08:00 |
|
hiyouga
|
679810a3d2
|
add codestral 22B
Former-commit-id: c23cc63d3d3c4fd8edd6c3b3ca1a2a32ec328d7d
|
2024-06-06 03:42:50 +08:00 |
|
hiyouga
|
8f25af89b6
|
lint
Former-commit-id: 7daf8366db0e161d46993fd87cf983a27a0ce2a3
|
2024-06-06 03:33:44 +08:00 |
|
hoshi-hiyouga
|
229794a148
|
Merge pull request #4066 from injet-zhou/main
add throughput entry to training log
Former-commit-id: f2580ad403cd0ae91aa0954c0a15363c46452438
|
2024-06-06 03:32:04 +08:00 |
|
hoshi-hiyouga
|
d31c9c73c7
|
Merge pull request #4080 from MengqingCao/npu
Add npu option for model exporting
Former-commit-id: ca459f67eb319c262d5bab882e9cb3535a76eb69
|
2024-06-06 03:15:44 +08:00 |
|