1172 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Jintao	3835dcc2c5	update (#1864)	2024-08-30 11:49:28 +08:00
Jintao	eb5b0a1ca5	update qwen2-vl docs (#1861)	2024-08-30 10:39:50 +08:00
Jintao Huang	de42edd669	fix docs	2024-08-30 01:18:30 +08:00
Jintao	b8dc24e04d	update qwen2-vl docs (#1858)	2024-08-30 01:11:55 +08:00
Jintao	f38ce0d8d3	update qwen2-vl docs (#1856)	2024-08-30 00:22:14 +08:00
tastelikefeet	6c7e4a268e	fix ui (#1855)	2024-08-29 22:13:21 +08:00
Jintao	ed755a51ba	support qwen2-vl & video finetune (#1849)	2024-08-29 20:36:03 +08:00
tastelikefeet	3b84266a5a	Support qwen2 vl grounding (#1854)	2024-08-29 20:23:34 +08:00
tastelikefeet	14f10bef45	Fix Pissa and OLoRA (#1852)	2024-08-29 19:19:53 +08:00
tastelikefeet	8a1606d49f	Fix some datasets for streaming (#1848)	2024-08-29 17:25:35 +08:00
tastelikefeet	a5338771cf	Add internvl2 awq models (#1846)	2024-08-29 16:20:21 +08:00
Jintao	17a32098a0	support qwen2-vl (#1842)	2024-08-29 15:32:57 +08:00
tastelikefeet	3a3c4dd332	Support eval_nproc (#1843)	2024-08-29 14:10:50 +08:00
jinghanhu	c2aeff1182	fix internlm-xcomposer rlhf (#1838) ... * fix internlm-xcomposer dpo * fix orpo/cpo	2024-08-28 14:15:54 +08:00
tastelikefeet	4ee93e6362	add ddp_timeout parameter (#1836) ... (cherry picked from commit f95307419c3eda83fceb9224994470463afe68de)	2024-08-27 21:29:50 +08:00
Jintao	dd923eb267	support qwen2-pro dataset (#1834)	2024-08-27 21:18:49 +08:00
tastelikefeet	8971d42b63	fix inject (#1835)	2024-08-27 21:07:15 +08:00
tastelikefeet	c4cbff9985	Fix code (#1824)	2024-08-27 18:53:51 +08:00
Jintao	68d5f6f092	fix minicpm-v 2.6 infer device_map (#1832)	2024-08-27 17:08:58 +08:00
Jintao	0d8575f55d	use default-lora (#1823)	2024-08-27 14:12:47 +08:00
Jintao	33e9935555	Support register loss func (#1822)	2024-08-26 21:44:05 +08:00
tastelikefeet	b950a9da24	fix dora deployment (#1821)	2024-08-26 18:03:10 +08:00
tastelikefeet	7a041d6c30	Support liger (#1819)	2024-08-26 16:29:01 +08:00
Jintao	58176fd5d3	fix preprocess_num_proc (#1818)	2024-08-26 16:28:35 +08:00
Jintao	044f15c04a	fix mp+ddp & resume_from_checkpoint (#1815)	2024-08-26 13:56:32 +08:00
Jintao	0a21505733	Support zero2 offload (#1814)	2024-08-26 11:03:24 +08:00
Jintao	a26f120735	compat with vllm0.5.5 (#1812)	2024-08-25 21:53:58 +08:00
tastelikefeet	e73c5e2396	fix (#1811)	2024-08-23 22:13:37 +08:00
王宁	0867c1d2d1	fix offline megatron export (#1805)	2024-08-23 21:59:48 +08:00
Jintao	603a655171	Support Latex-OCR dataset (#1810)	2024-08-23 21:19:34 +08:00
Jintao Huang	ebc0a90d8e	update docs link	2024-08-23 14:29:17 +08:00
Jintao	e29cf5a875	Support hd_num (#1801)	2024-08-23 14:14:59 +08:00
王宁	089234c71e	fix megatron_patch_path (#1804)	2024-08-23 14:06:57 +08:00
tastelikefeet	1a84728136	fix citest (#1797)	2024-08-23 10:00:27 +08:00
jinghanhu	be208fac2b	fix mllm rlhf with full sft type (#1800) ... * fix * fix	2024-08-22 23:33:25 +08:00
Jintao	422ff7bb2b	fix history_roles (#1798)	2024-08-22 21:55:09 +08:00
tastelikefeet	2ff689e2b9	fix imports (#1796)	2024-08-22 21:06:09 +08:00
Jintao	3d5a07cb8c	fix stream bugs (#1794)	2024-08-22 19:32:55 +08:00
Jintao	ab8476ca6f	fix yi-vl template (#1793)	2024-08-22 16:12:51 +08:00
Jintao	af3da08b78	support qwen-vl & base64 (#1790)	2024-08-22 15:07:07 +08:00
tastelikefeet	2e47ded6c0	update doc (#1789)	2024-08-22 10:56:27 +08:00
tastelikefeet	b940c740f7	ReFT (#1785)	2024-08-22 00:59:04 +08:00
Jintao	63cb9de54d	support phi3.5-vision (#1780)	2024-08-21 20:47:13 +08:00
Jintao	b8f0268030	fix moe & gradient_checkpointing (#1782)	2024-08-21 18:49:08 +08:00
Jintao	c7accb476c	fix dataset_test_ratio (#1779)	2024-08-21 11:46:41 +08:00
Jintao	460a97876e	Fix zero3 & minicpm-v/internvl2/xcomposer (#1772)	2024-08-21 01:10:51 +08:00
Jintao	2fe7c224b0	Fix qwen2-audio & zero3 (#1774)	2024-08-20 16:56:59 +08:00
Jintao	554418c5c4	Fix deepseek-coder-v2-lite template (#1771)	2024-08-20 14:30:32 +08:00
Baole Ai	cb2460884d	[TorchAcc] fix: fix save/load checkpoint for full sft FSDP (#1765)	2024-08-20 13:33:00 +08:00
Jintao	e88b62d2e0	Support llava onevision (#1761)	2024-08-20 11:10:34 +08:00