47 Commits

Author SHA1 Message Date
hiyouga
0385c60177 fix #6391
Former-commit-id: 067ba6e6cb4d8a1d95bba0a108f73008416a2865
2024-12-19 12:16:38 +00:00
hiyouga
01eeae50b5 support disable shuffling
Former-commit-id: 9d8c35fd6b838ede0bd6827c6c6121f2cba2b11b
2024-12-19 08:53:21 +00:00
hiyouga
19ebc0e7a2 support control eos, fix #6345
Former-commit-id: cb0f8399356bf372f3b7963f2565c3d504be0923
2024-12-17 10:42:05 +00:00
hiyouga
aacd9642f5 fix #6348
Former-commit-id: 83e552320909f4775377889f1512994b7e638a7e
2024-12-17 10:06:46 +00:00
hiyouga
fb22651faf fix mrope
Former-commit-id: 55bee1d333549ca19858b3f5c1b7b86926e5fb09
2024-12-12 15:08:17 +00:00
hiyouga
c1768cfb14 support batch infer in vllm
Former-commit-id: 3ef5ed3b9a44eed2f7e3ff221dfc343d0a97c0b5
2024-12-04 13:50:00 +00:00
Ting
87b1f851f1 code refactor
Former-commit-id: ee3f85aa9677d0aeecb3bc396530d2cd7c50dce5
2024-11-19 20:33:18 +08:00
Ting
a20c2b6ecf update
Former-commit-id: a3e8ca53e654136242197a2da872cc0e5cf67880
2024-11-19 19:10:07 +08:00
Ting
fee94e1c54 support efficient tokens calculation on sft/dpo
Former-commit-id: b157d5cccdeb42412b8b440d25d5bdfa8a50be68
2024-11-19 17:15:47 +08:00
hiyouga
2bb3255e74 fix dpo metrics
Former-commit-id: 57029280da825a39fbf5a05097921b861f126669
2024-11-02 20:59:01 +08:00
hiyouga
093eda2ad6 support rank0 logger
Former-commit-id: 84528eabe560091bfd866b6a0ca864085af7529b
2024-11-02 18:31:04 +08:00
hiyouga
8185eb1890 fix incorrect loss value for vlms
Former-commit-id: 0aa29a71ce958343a2086090d647eb63b8f5f5be
2024-10-30 08:56:46 +00:00
hiyouga
7f71276ad8 add docstrings, refactor logger
Former-commit-id: c34e489d71f8f539028543ccf8ee92cecedd6276
2024-09-08 00:56:56 +08:00
hiyouga
af178cbcd1 update get template
Former-commit-id: 21ea0d0786f91c0bce79630963e66b815a6792a0
2024-09-04 22:36:20 +08:00
hiyouga
7056087e92 lazy image load
Former-commit-id: cdd733b575411e003bc5ffd6560dd8eff8aa09cf
2024-09-04 02:27:08 +08:00
hiyouga
2e1396cd6b lint
Former-commit-id: d821d933e6cb982d648a69f85f6ad01d0560ed70
2024-09-03 00:46:25 +08:00
hiyouga
b5e9df5df8 fix #5324
Former-commit-id: f7aa06c9c0b18c28419ea5792410915d3f322cbf
2024-09-02 23:56:21 +08:00
hoshi-hiyouga
7367c6ec21 fix trainer predict
Former-commit-id: 2790790cd26c6743105555a60523b89f367ebce3
2024-09-02 10:15:29 +08:00
hoshi-hiyouga
6579ec8c4c remove .cpu()
Former-commit-id: 35c57cc9dcba305d40282a9757ddc23968c210ac
2024-09-02 10:10:53 +08:00
hiyouga
2f6fc27c8b remove visual_inputs, fix qlora
Former-commit-id: be30c01c4f1482520ece770bd54c6a4837c26f0a
2024-08-31 00:24:51 +08:00
hiyouga
d789b667d7 optimize predict vram
Former-commit-id: a577e44eee351b3ed8011a33ae01cd713354ff97
2024-08-30 23:08:45 +08:00
moontidef
33a90b9026 fix: rename optimzer to optimizer
Former-commit-id: 186dc1fde822e6a603ac273538741ea3853f243e
2024-08-07 10:05:01 +08:00
hiyouga
341225a405 fix metrics #4786
Former-commit-id: 7d0c4bd394fc3cba197db1719f1164b9dd66ac21
2024-07-17 00:47:00 +08:00
hiyouga
8c93921952 support batch_eval_metrics, fix #4826
Former-commit-id: 3fe1df17188825f8a32fbe6a1294b4b532ce0c85
2024-07-17 00:33:00 +08:00
hiyouga
1891b64072 fix #4820
Former-commit-id: 8c0f8357e1eebee32010fe715554f1136b68b4ba
2024-07-15 22:32:07 +08:00
hiyouga
e4d11a117b fix up
Former-commit-id: 43a56cb331fae899ca35b0c312730d4ab79d0c42
2024-07-15 01:04:56 +08:00
codingma
5f2bd04799 1. add custom eval dataset support
2. merge load dataset and split dataset function


Former-commit-id: 963d97ba07e7efa3a4544c4d077283d9e112b3ad
2024-07-05 15:52:10 +08:00
hiyouga
3d219b91b9 fix packing for eager/sdpa attn
Former-commit-id: 735a033ceb7f2da6da71d138ea091d8a665411a9
2024-07-04 01:52:43 +08:00
hiyouga
e6ba7ef3e6 improve rlhf
Former-commit-id: e441780e3db256ca09a442ea9254e7ce16898a07
2024-07-02 22:23:08 +08:00
hiyouga
973cf8e980 tiny fix
Former-commit-id: 5dd2e5c3323f56420b5845a5ed28bcd9d4da5e41
2024-07-01 05:43:17 +08:00
hiyouga
884b49e662 add eval acc
Former-commit-id: 7ffde76fbfb6192e3aac31ccc098f31ce89181ae
2024-07-01 03:51:20 +08:00
hiyouga
46f0189e88 refactor pissa, improve llamaboard
Former-commit-id: 619556e46c19718f702c97df5d570a2a4c5fb13a
2024-06-28 01:04:24 +08:00
hzhaoy
89d9dd5aa5 fix #4579
Former-commit-id: 0fa298ff6a4febea36ea9f11c7594277a77e6e9b
2024-06-27 13:49:57 +08:00
hiyouga
9fd7a410bb tiny fix about badam
Former-commit-id: 03f49267c7406e36aee35639f86e6e0383897090
2024-06-25 01:54:53 +08:00
Jonery
fa3150548e Cleaner integration.
Former-commit-id: 26d4b05d424bd71f570195dd433258caf6465d92
2024-06-19 12:29:40 +08:00
Jonery
12fcfc2b72 Support distributed BAdam.
Former-commit-id: bdcb986e37975911c190a74d3e60bb77aa2033bd
2024-06-18 12:27:47 +08:00
Jonery
95ae30f678 Merge remote-tracking branch 'upstream/main'
Former-commit-id: 37834a7e79473ccf50ad7f67745b97c274c326d9
2024-06-17 18:44:51 +08:00
Jonery
ba303fd1aa adapt for badam with ds zero3
Former-commit-id: fff2a020ec8713022bd8145f4a7168168ea07ca4
2024-06-17 18:18:10 +08:00
hiyouga
32f45c9e91 support pissa
Former-commit-id: ef8e45f2eaf466c54e9a671512a2974575677b08
2024-06-16 01:08:12 +08:00
hiyouga
05f3a3c944 tiny fix
Former-commit-id: f7f440986b0ae3b38ea9f2da80789629d4f79ea1
2024-06-16 01:06:41 +08:00
hiyouga
bb88536166 add license
Former-commit-id: 69cfc98d7c81756a5ab6bf962240e393e449fef0
2024-06-15 17:54:33 +08:00
hiyouga
a30931fe0f fix #4295
Former-commit-id: 08f657868f9d605b837c5d8c2946a25cc05c8735
2024-06-15 04:34:55 +08:00
hiyouga
49b58fd6af fix #4221
Former-commit-id: 05a3be4853b941909e7d193c31e8d62c8c5f879b
2024-06-13 02:48:21 +08:00
hiyouga
0a75224f62 clean code
Former-commit-id: f54cafd5c7f0383370d1a2f357834a61a97397ce
2024-06-13 01:58:16 +08:00
hiyouga
b0e5a76f4c fix ppo trainer save zero3 model
accelerator.get_state_dict(ds_model) should be called at all ranks


Former-commit-id: 3a0f60f0aa072531e4ae5819ec00c8fa42aa0913
2024-06-07 05:14:19 +08:00
hiyouga
fcb134e144 rename files
Former-commit-id: e1a8431770fc36c0c9ee7fed4abbc3d7fdcc5efd
2024-06-07 00:09:06 +08:00
hiyouga
dfa686b617 rename package
Former-commit-id: a07ff0c083558cfe6f474d13027642d3052fee08
2024-05-16 18:39:08 +08:00