4863 commits

This branch

This branch

All branches

Author	SHA1	Message	Date
oobabooga	cc964ee579	mtmd: Increase the size of the UI image preview	2025-08-10 07:44:38 -07:00
oobabooga	6fbf162d71	Default max_tokens to 512 in the API instead of 16	2025-08-10 07:21:55 -07:00
oobabooga	1fb5807859	mtmd: Fix API text completion when no images are sent	2025-08-10 06:54:44 -07:00
oobabooga	0ea62d88f6	mtmd: Fix "continue" when an image is present	2025-08-09 21:47:02 -07:00
oobabooga	4663b1a56e	Update docs	2025-08-09 21:45:50 -07:00
oobabooga	2f90ac9880	Move the new image_utils.py file to modules/	2025-08-09 21:41:38 -07:00
oobabooga	c6b4d1e87f	Fix the exllamav2 loader ignoring add_bos	2025-08-09 21:34:35 -07:00
oobabooga	d86b0ec010	Add multimodal support (llama.cpp) (#7027)	2025-08-10 01:27:25 -03:00
oobabooga	eb16f64017	Update llama.cpp	2025-08-09 17:12:16 -07:00
oobabooga	a289a92b94	Fix exllamav3 token count	2025-08-09 17:10:58 -07:00
oobabooga	d489eb589a	Attempt at fixing new exllamav3 loader undefined behavior when switching conversations	2025-08-09 14:11:31 -07:00
oobabooga	a6d6bee88c	Change a comment	2025-08-09 07:51:03 -07:00
oobabooga	2fe79a93cc	mtmd: Handle another case after 3f5ec9644f	2025-08-09 07:50:24 -07:00
oobabooga	59c6138e98	Remove a log message	2025-08-09 07:32:15 -07:00
oobabooga	f396b82a4f	mtmd: Better way to detect if an EXL3 model is multimodal	2025-08-09 07:31:36 -07:00
oobabooga	fa9be444fa	Use ExLlamav3 instead of ExLlamav3_HF by default for EXL3 models	2025-08-09 07:26:59 -07:00
oobabooga	d9db8f63a7	mtmd: Simplifications	2025-08-09 07:25:42 -07:00
oobabooga	3f5ec9644f	mtmd: Place the image <__media__> at the top of the prompt	2025-08-09 07:06:07 -07:00
oobabooga	1168004067	Minor change	2025-08-09 07:01:55 -07:00
oobabooga	9e260332cc	Remove some unnecessary code	2025-08-08 21:22:47 -07:00
oobabooga	544c3a7c9f	Polish the new exllamav3 loader	2025-08-08 21:15:53 -07:00
oobabooga	8fcadff8d3	mtmd: Use the base64 attachment for the UI preview instead of the file	2025-08-08 20:13:54 -07:00
oobabooga	6e9de75727	Support loading chat templates from chat_template.json files	2025-08-08 19:35:09 -07:00
Katehuuh	88127f46c1	Add multimodal support (ExLlamaV3) (#7174)	2025-08-08 23:31:16 -03:00
oobabooga	b391ac8eb1	Fix getting the ctx-size for EXL3/EXL2/Transformers models	2025-08-08 18:11:45 -07:00
oobabooga	f1147c9926	Update llama.cpp	2025-08-06 19:32:36 -07:00
oobabooga	3e24f455c8	Fix continue for GPT-OSS (hopefully the final fix)	2025-08-06 10:18:42 -07:00
oobabooga	0c1403f2c7	Handle GPT-OSS as a special case when continuing	2025-08-06 08:05:37 -07:00
oobabooga	6ce4b353c4	Fix the GPT-OSS template	2025-08-06 07:12:39 -07:00
oobabooga	7c82d65a9d	Handle GPT-OSS as a special template case	2025-08-05 18:05:09 -07:00
oobabooga	fbea21a1f1	Only use enable_thinking if the template supports it	2025-08-05 17:33:27 -07:00
oobabooga	bfbbfc2361	Ignore add_generation_prompt in GPT-OSS	2025-08-05 17:33:01 -07:00
oobabooga	20adc3c967	Start over new template handling (to avoid overcomplicating)	2025-08-05 16:58:45 -07:00
oobabooga	80f6abb07e	Begin fixing 'Continue' with GPT-OSS	2025-08-05 16:01:19 -07:00
oobabooga	e5b8d4d072	Fix a typo	2025-08-05 15:52:56 -07:00
oobabooga	701048cf33	Try to avoid breaking jinja2 parsing for older models	2025-08-05 15:51:24 -07:00
oobabooga	7d98ca6195	Make web search functional with thinking models	2025-08-05 15:44:33 -07:00
oobabooga	0e42575c57	Fix thinking block parsing for GPT-OSS under llama.cpp	2025-08-05 15:36:20 -07:00
oobabooga	498778b8ac	Add a new 'Reasoning effort' UI element	2025-08-05 15:19:11 -07:00
oobabooga	6bb8212731	Fix thinking block rendering for GPT-OSS	2025-08-05 15:06:22 -07:00
oobabooga	42e3a7a5ae	Update llama.cpp	2025-08-05 14:56:12 -07:00
oobabooga	5c5a4dfc14	Fix impersonate	2025-08-05 13:04:10 -07:00
oobabooga	ecd16d6bf9	Automatically set skip_special_tokens to False for channel-based templates	2025-08-05 12:57:49 -07:00
oobabooga	178c3e75cc	Handle templates with channels separately	2025-08-05 12:52:17 -07:00
oobabooga	9f28f53cfc	Better parsing of the gpt-oss template	2025-08-05 11:56:00 -07:00
oobabooga	3b28dc1821	Don't pass torch_dtype to transformers loader, let it be autodetected	2025-08-05 11:35:53 -07:00
oobabooga	3039aeffeb	Fix parsing the gpt-oss-20b template	2025-08-05 11:35:17 -07:00
oobabooga	5989043537	Transformers: Support standalone .jinja chat templates (for GPT-OSS)	2025-08-05 11:22:18 -07:00
oobabooga	02a3420a50	Bump transformers to 4.55 (adds gpt-oss support)	2025-08-05 10:09:30 -07:00
oobabooga	74230f559a	Bump transformers to 4.54	2025-08-01 11:03:15 -07:00