32 Commits

Author	SHA1	Message	Date
Deyao Zhu	8c2297750f	update env name to minigptv, add bug report and feature request back	2023-10-23 20:42:36 +03:00
Deyao Zhu	1e8e62a889	remove pycache in git	2023-10-23 20:18:21 +03:00
Deyao Zhu	0918b0314c	remove pycache in git	2023-10-23 20:12:59 +03:00
junchen14	ba636c699c	add wandb	2023-10-23 09:13:48 +03:00
junchen14	5c3ec8bb73	Merge branch 'main' of https://github.com/junchen14/MiniGPT-4_finetune into main	2023-10-23 07:08:42 +03:00
junchen14	9005196851	add datasets	2023-10-23 07:05:27 +03:00
Deyao Zhu	ba6cdb6992	Merge branch 'main' of github.com:TsuTikgiau/MiniGPT-4 into main	2023-10-22 22:32:01 +03:00
Deyao Zhu	07961c1a56	fix the demo in conversation template	2023-10-22 22:29:58 +03:00
junchen14	0e5d34ad2e	add finetuning code	2023-10-22 21:37:45 +03:00
Deyao Zhu	89b6fbfb78	update the batch size code	2023-10-21 22:06:29 +03:00
Deyao Zhu	99383a919f	fix the chat template and reduction error in v1 training	2023-10-21 16:32:54 +03:00
Deyao Zhu	468b452a8a	fix the demo error when temperature = 2	2023-10-16 16:01:20 +03:00
Deyao Zhu	a3709f52e0	reset the path in config files	2023-10-13 17:11:11 +03:00
Deyao Zhu	0eba23ce3b	update v2 demo	2023-10-13 03:14:35 +03:00
Deyao Zhu	7a575af639	init version of v2	2023-10-13 00:14:34 +03:00
Deyao Zhu	d1367e5e64	rename mini_gpt4.py to minigpt_4.py and remove some unused import	2023-10-12 22:47:48 +03:00
Deyao Zhu	045a1d0602	modularize minigpt4 code	2023-10-12 22:44:49 +03:00
Deyao Zhu	90b7b00268	update minigpt_base.py expect the initailization function	2023-10-12 17:17:32 +03:00
Deyao Zhu	4ce9c5febe	update modeling_llama.py	2023-10-12 15:03:56 +03:00
Deyao Zhu	062ad9bb30	remove files	2023-10-12 14:57:17 +03:00
Deyao Zhu	871918f758	fix the device bug with latest transformers	2023-08-29 16:12:22 +03:00
Deyao Zhu	fb8e2c656a	include llama2	2023-08-28 21:26:00 +03:00
Jun Chen	3bd99950f0	Update runner_base.py	2023-04-23 15:49:37 +03:00
Deyao Zhu	322ed189e6	add argument to set the demo device	2023-04-19 19:49:05 +03:00
Deyao Zhu	f5c28362d4	fix the issue with strange strikethrough output	2023-04-19 10:53:33 +03:00
Deyao Zhu	dadc0d7e69	adding length control and change the default hyperparameter of the conversation to avoid OOM in 3090.	2023-04-18 22:04:50 +03:00
Deyao Zhu	3c9d321b79	fix an import bug in image_text_pair_builder.py	2023-04-17 16:57:20 +03:00
Deyao Zhu	b76d5c5a50	add argument to switch 8bit	2023-04-17 16:54:00 +03:00
152334H	700f05d079	set num_beam default to 1 instead of completely removing it	2023-04-17 14:56:18 +08:00
152334H	6a8e6e49ff	load in 8bit and run ViT-L in cpu to reduce vram use	2023-04-17 13:52:09 +08:00
152334H	017fa43d03	disable beam search to reduce vram use	2023-04-17 13:51:39 +08:00
Deyao Zhu	f1a33af227	first commit	2023-04-17 01:04:16 +03:00