612 Commits

Author SHA1 Message Date
hiyouga
61960189b2 fix #1789
Former-commit-id: 4571068e1e00dc234c9131185fe0924c726add84
2024-01-09 18:31:27 +08:00
hiyouga
f53bc7d9a0 fix #2127
Former-commit-id: ebee4f6a2a20a39ad7fbb87edfa63e244da7b5e6
2024-01-09 14:49:13 +08:00
hiyouga
de9148930f fix #2125
Former-commit-id: 3ae735ffe8acbe1df05324daa9e18ec37f33d594
2024-01-08 21:42:25 +08:00
hiyouga
5b61063048 fix api server
Former-commit-id: 08464183b9b034abdbf179d7043705a0754837e5
2024-01-07 17:14:42 +08:00
hiyouga
6bbcf5ad16 improve model export
Former-commit-id: d2a676c8ba550e1dd7f4e12cb397a32e01831d85
2024-01-05 18:51:49 +08:00
hiyouga
7c50959fe6 fix #2098
Former-commit-id: f6fdd83f8a6bf3e48cf08fd098ec7b08d34d16d7
2024-01-05 17:11:26 +08:00
hiyouga
a78cd5c5ad fix qwen template
Former-commit-id: ed216bbc4693860eaf27c2a95108f705d530cbdb
2024-01-05 16:14:56 +08:00
hiyouga
dc8714a003 fix #2081
Former-commit-id: 33f2c0d4f89cf76671c0fdfbcee79d732b6a020e
2024-01-04 23:19:08 +08:00
hiyouga
dabd40750c fix #2090
Former-commit-id: cc275abe09665668289eb11efb446d1876318110
2024-01-04 23:05:08 +08:00
hiyouga
9a496950aa fix #2067
Former-commit-id: 368b31f6b7422562e4cd471d54affe644c4f10cb
2024-01-04 22:53:03 +08:00
hiyouga
3fff67e3c7 fix dispatch
Former-commit-id: 1696698eb980fef36a44f860105711bc06d674c9
2024-01-03 16:33:16 +08:00
hiyouga
eb021ca748 fix valuehead patch
Former-commit-id: 24d8d6f224ccb98387ec72e688fa32f5f308dd07
2024-01-03 16:19:23 +08:00
hiyouga
1a86cc3078 fix rm server
Former-commit-id: 55021097d565536a68113ee33af31beaff38334e
2024-01-03 15:30:46 +08:00
hiyouga
c14bf033a3 fix #2014
Former-commit-id: ce2156eaa83e2b572f4198a1eb77a835fb502319
2023-12-29 15:17:22 +08:00
hiyouga
4735cb96c1 add yuan model
Former-commit-id: c7ea17d6168d0a032960dbf51c12482f97529b1e
2023-12-29 13:50:24 +08:00
hiyouga
afa3c86391 fix version
Former-commit-id: 47da742fc971520a697ed5ce611f8d6de61ab22b
2023-12-29 04:53:36 +08:00
hiyouga
cae66bce3d fix args
Former-commit-id: 65c5b0477c0e62691a1f8790670ba04d7f6d2804
2023-12-28 18:47:19 +08:00
hiyouga
16688b773a fix export format
Former-commit-id: e165354facf7e69f535f9b7d99438f03dbf0293d
2023-12-28 18:40:46 +08:00
hiyouga
d0946f08db fix ppo trainer
Former-commit-id: 5431be42f9c43095d478f2250fac64ef189eb3ad
2023-12-28 18:09:28 +08:00
hiyouga
498b83d77a fix bug
Former-commit-id: e4bb846c43c087bfdd99d0c9eb0318e95b943977
2023-12-24 19:20:12 +08:00
hiyouga
2a3980d6ba update loader
Former-commit-id: 6629087e12f64f2635f24311234202077814083c
2023-12-24 19:10:23 +08:00
hiyouga
5d440f978e update patcher
Former-commit-id: e44b82ee245a7ee99057c7b58b1edef5c222dc1f
2023-12-23 15:24:27 +08:00
hiyouga
11a6c8a9a0 fix #1909
Former-commit-id: 0bbf7118dfeb936baf03e74269211fa593003cfe
2023-12-23 14:42:20 +08:00
hiyouga
a7fc20cb2d fix unsloth dtype
Former-commit-id: 779cfefb7841d00fc712a5f5addf0fe3eb14c6fd
2023-12-23 01:59:49 +08:00
hiyouga
938c4cb132 fix dpo trainer
Former-commit-id: 074745b1707f98e092749f57041d866c5d55bc04
2023-12-23 01:51:55 +08:00
hiyouga
bf872424da llama board: add unsloth
Former-commit-id: 9a18a85639ad53d75fb25b6a71edd96fe95c5e59
2023-12-23 00:35:53 +08:00
hiyouga
f0d405f392 support unsloth
Former-commit-id: 7aad0b889d9a316fffd65f32a419078418fc0986
2023-12-23 00:14:33 +08:00
ShaneTian
3eaabe12aa Fix slow model initialization in bfloat16 dtype.
Former-commit-id: d032daa4bd598dd0d71b43eb68a614de77a699a6
2023-12-22 16:27:28 +08:00
hiyouga
ce79528bb1 fix param type
Former-commit-id: ba69378841778410f8004385df3fd4c41e5fa573
2023-12-21 17:33:01 +08:00
hiyouga
fee0fef052 fix ds zero3 check
Former-commit-id: 083355fc051f5d25400eb80887ff5e0d15ce729b
2023-12-21 01:19:22 +08:00
hiyouga
de6be321b9 match version
Former-commit-id: af0194e6d9ba1c67a8592c840c01cee7991b0b0e
2023-12-20 22:17:35 +08:00
hiyouga
9337568e23 fix stop words
Former-commit-id: dec360d5aee58c11804b8e45dd8d4c375086887f
2023-12-20 19:06:43 +08:00
hiyouga
622d31e398 fix yi template #1895
Former-commit-id: 5af8841c4f6c97df522d2cf4e283d5ef0af21a18
2023-12-20 18:58:16 +08:00
hiyouga
23a875a8b1 improve quantization
Former-commit-id: 624cc212819b7cd16295c72084cd454b67cf89a6
2023-12-20 18:27:16 +08:00
hiyouga
c1233ab65f add max_memory for gptq #1923
Former-commit-id: c4a3977ad7278b59a9b3dadcb446bb4c99da5c9d
2023-12-20 18:15:17 +08:00
hiyouga
82a79e9fdf fix #1073 #1462 #1735 #1908
Former-commit-id: 31165a9822bd52130b33cd3439f887c26e0679dc
2023-12-20 17:15:40 +08:00
hiyouga
f64be8ee84 optimize data loading logic
Former-commit-id: ec1fe1daa98c61f62c753b22847de028b5c5cded
2023-12-20 16:15:41 +08:00
hiyouga
633624dc3c fix #1909
Former-commit-id: c6abbbfe90dcb0e832f73f0c611fc32eaa7ea78d
2023-12-20 16:11:07 +08:00
hiyouga
a862ce636f fix mixtral inference #1821
Former-commit-id: f86857bd9ef456e77ad79a584f1fa08a129e5270
2023-12-20 15:11:15 +08:00
hiyouga
e06b9c4fa1 fix #1900
Former-commit-id: 0c6ab7c75e671348f24d28309d450aecec0a157f
2023-12-19 17:21:46 +08:00
hiyouga
3981927d6b add codegeex template
Former-commit-id: a67a440644687dc2262134c0f2895f3ae42cae19
2023-12-18 19:52:35 +08:00
hiyouga
51c636db54 add xverse-65B-2 model
Former-commit-id: 2df923540c3cbf3b06c74801ea66d3523718b84a
2023-12-18 19:24:09 +08:00
hiyouga
1af13cb737 add models
Former-commit-id: 709ac8870a17a96e786b32c75ad8c4e573148cee
2023-12-18 19:09:31 +08:00
hiyouga
5a199af387 fix tokenizer for Yi chat models #1617 #1875
Former-commit-id: 71a9c1617181b7df46cfb193464fb7e56e6399b1
2023-12-18 17:18:11 +08:00
hiyouga
16cc0321f2 fix llama board
Former-commit-id: c46879575f434b2b458bddae6db63b227db4202e
2023-12-16 22:17:37 +08:00
hiyouga
8154b4bdf6 fix #1742
Former-commit-id: 870426ff70c060213ac283b10a9b1f4bf71679ef
2023-12-16 20:50:45 +08:00
hiyouga
397f6bb615 add xverse-65b-chat model
Former-commit-id: 7ae6919b9bb9ecc8d821eea47a03eacd9eb997ac
2023-12-16 20:21:29 +08:00
hiyouga
ff5f57bfbf set version
Former-commit-id: 328ad06bd4da01d6534354eb6bb798b259a017b9
2023-12-16 20:17:51 +08:00
hiyouga
7c362509a6 add noisy mean initialization #1815
Former-commit-id: a66186b8724ffd0351a32593ab52d8a2312f339b
2023-12-16 19:47:51 +08:00
hiyouga
4e75ca1222 support dpo-ftx
Former-commit-id: b87c74289d523ef88611b376074199ffd03cf103
2023-12-16 19:21:41 +08:00