1172 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
Xiulong Yuan	6b30748c3c	support batch flattening collator (#2499)	2024-11-25 17:04:12 +08:00
Jintao	5431a57928	fix eval_dataset no (#2497)	2024-11-25 13:47:31 +08:00
Jintao	279b524314	Support marco o1 (#2496)	2024-11-25 11:31:35 +08:00
Jintao	4f7054fb24	Fix preprocess num proc (#2492)	2024-11-23 18:09:47 +08:00
Jintao	b2e5d6e677	fix awq quant device_map (#2488)	2024-11-22 17:51:21 +08:00
slin000111	cf574d99bd	Update Common QA (#2475)	2024-11-20 10:47:59 +08:00
Jintao	2659303869	fix kto (#2478)	2024-11-19 19:26:01 +08:00
Xiulong Yuan	552190740c	update padding strategy for persistent cache (#2464)	2024-11-18 15:28:58 +08:00
Jintao	616480e301	fix qwen2vl pt infer (#2463)	2024-11-18 10:26:18 +08:00
Jintao	b6c0a58cad	fix peft is_multimodal (#2462)	2024-11-16 23:41:41 +08:00
Jintao	10a1143a10	fix qwen2-vl position_ids (#2461)	2024-11-16 20:36:21 +08:00
tastelikefeet	5c229180e5	remove useless code (#2453)	2024-11-14 21:37:06 +08:00
tastelikefeet	78b6f78155	bump ms version (#2449)	2024-11-13 22:39:36 +08:00
tastelikefeet	490349a676	support part tuner replace_key False (#2438)	2024-11-13 17:37:20 +08:00
Jintao	d13c431a0f	fix internvl2 lmdeploy>=0.6 deploy (#2429)	2024-11-11 20:55:46 +08:00
Jintao	b0700b33e1	compat transformers==4.47 (#2426)	2024-11-11 17:06:17 +08:00
Jintao	2075824675	fix docs (#2425)	2024-11-11 15:17:59 +08:00
jinghanhu	1e878b9ca1	fix PPO (#2377)	2024-11-11 14:19:45 +08:00
Jintao	b095d7c2b2	Support qwen2 5 coder series model (#2422)	2024-11-11 13:53:43 +08:00
Jintao	f95e19ce05	Fix dataset map (#2421)	2024-11-11 12:35:57 +08:00
Jintao	acd17e5a7d	Fix molmo infer (#2419)	2024-11-09 15:08:45 +08:00
Jintao	93032f418e	Compat transformers 4.46.2 loss (#2413)	2024-11-08 13:24:16 +08:00
Jintao	8324fdac1b	fix qwen_vl dpo (#2411)	2024-11-08 11:45:21 +08:00
Jintao Huang	a57d1c6bfe	fix qwen_vl lmdeploy	2024-11-08 11:17:44 +08:00
Yingda Chen	53790bd87f	fix model path (#2410) ... Co-authored-by: Yingda Chen <yingda.chen@alibaba-inc.com>	2024-11-08 09:44:23 +08:00
Yingda Chen	f5f6eafb23	Revert "fix model path" ... This reverts commit a034fca3c6.	2024-11-08 08:54:23 +08:00
Yingda Chen	a034fca3c6	fix model path	2024-11-08 08:44:16 +08:00
Jintao	6a307bd89d	fix qwen_vl npu (#2408)	2024-11-07 21:05:42 +08:00
Jintao	1fc6e3def2	support qwen2.5-coder (#2400)	2024-11-07 10:34:17 +08:00
Jintao	8b1219deb9	fix lmdeploy warning & gptq-int4 support (#2401)	2024-11-07 10:25:37 +08:00
Jintao	3465f13423	fix swift deploy (lmdeploy stream) (#2397)	2024-11-06 19:29:43 +08:00
Jintao	ad65d9e4df	fix_mplug_owl3_cut_shape (#2394)	2024-11-06 14:42:03 +08:00
Jintao	2952781e43	fix deploy stream media_type (#2393)	2024-11-06 14:07:28 +08:00
Jintao	4d0d902de4	fix trl transformers==4.46 compat (#2385)	2024-11-05 19:58:13 +08:00
Baole Ai	fb808bdcc1	[TorchAcc] fix qwen2 for transformers>=4.45 (#2379)	2024-11-05 10:04:37 +08:00
Jintao	7dd9ae01c0	fix gptq ovis quant (inputs_embeds) (#2378)	2024-11-04 13:10:30 +08:00
Jintao	e94e1e46cd	fix custom loss (#2374)	2024-11-01 22:00:15 +08:00
DaozeZhang	6b886704c6	Support Ministral-8b-inst-2410 (#2371)	2024-10-31 16:50:53 +08:00
Jintao	44302aba29	fix transformers>=4.46 loss (#2365)	2024-10-31 10:32:02 +08:00
Jintao	59f58c3087	fix internvl2 eval (#2362)	2024-10-30 22:03:32 +08:00
tastelikefeet	f4e09d8516	fix doc (#2355)	2024-10-29 21:54:44 +08:00
tastelikefeet	6b33108084	fix streaming (#2350)	2024-10-28 15:55:57 +08:00
Jintao	280edc312d	compat transformers 4.46 (#2341)	2024-10-25 18:06:13 +08:00
Aunali321	825ec455b5	feat: Add aya models (#2335)	2024-10-25 17:29:20 +08:00
Jintao	ef43783745	fix enable_cache (#2332)	2024-10-24 18:55:51 +08:00
Yunlin Mao	8bf0bb4e69	fix custom vllm eval args (#2325) ... * fix custom vllm eval args * fix lint * fix eval.py * update --------- Co-authored-by: Qing112 <autumnwindji@foxmail.com>	2024-10-23 20:45:07 +08:00
DaozeZhang	c87c5ee17f	Support Moviechat-1k-test dataset, a video datatset (#2319) ... * add transformers in gitignore * fix a typo bug in text-caps * add .run into gitignore * add vlmeval to gitignore * add my_model/ to gitignore * support moviechat-1l-test dataset * support movie-chat-1k datset, and add doc * modify var name to is_not_compressed_file * update	2024-10-23 15:49:14 +08:00
Hong Zhang	6ef3ec7014	support emu3-chat (#2322)	2024-10-23 15:49:00 +08:00
Jintao	d42f8b5c3d	fix molmo device_map (#2311)	2024-10-22 15:22:53 +08:00
Hong Zhang	806cedd8b5	update molmo readme (#2313)	2024-10-22 15:15:29 +08:00