146 commits

Author	SHA1	Message	Date
oobabooga	9ec46b8c44	Remove the HQQ loader (HQQ models can be loaded through Transformers)	2025-05-19 09:23:24 -07:00
oobabooga	93e1850a2c	Only show the VRAM info for llama.cpp	2025-05-15 21:42:15 -07:00
oobabooga	3fa1a899ae	UI: Fix gpu-layers being ignored (closes #6973)	2025-05-13 12:07:59 -07:00
oobabooga	a2ab42d390	UI: Remove the exllamav2 info message	2025-05-08 08:00:38 -07:00
oobabooga	c4f36db0d8	llama.cpp: remove tfs (it doesn't get used)	2025-05-06 08:41:13 -07:00
oobabooga	1927afe894	Fix top_n_sigma not showing for llama.cpp	2025-05-06 08:18:49 -07:00
oobabooga	d10bded7f8	UI: Add an enable_thinking option to enable/disable Qwen3 thinking	2025-04-28 22:37:01 -07:00
oobabooga	4a32e1f80c	UI: show draft_max for ExLlamaV2	2025-04-26 18:01:44 -07:00
oobabooga	d4017fbb6d	ExLlamaV3: Add kv cache quantization (#6903)	2025-04-25 21:32:00 -03:00
oobabooga	d4b1e31c49	Use --ctx-size to specify the context size for all loaders ... Old flags are still recognized as alternatives.	2025-04-25 16:59:03 -07:00
oobabooga	98f4c694b9	llama.cpp: Add --extra-flags parameter for passing additional flags to llama-server	2025-04-25 07:32:51 -07:00
oobabooga	ae1fe87365	ExLlamaV2: Add speculative decoding (#6899)	2025-04-25 00:11:04 -03:00
oobabooga	e99c20bcb0	llama.cpp: Add speculative decoding (#6891)	2025-04-23 20:10:16 -03:00
oobabooga	15989c2ed8	Make llama.cpp the default loader	2025-04-21 16:36:35 -07:00
oobabooga	ae02ffc605	Refactor the transformers loader (#6859)	2025-04-20 13:33:47 -03:00
oobabooga	8144e1031e	Remove deprecated command-line flags	2025-04-18 06:02:28 -07:00
oobabooga	ae54d8faaa	New llama.cpp loader (#6846)	2025-04-18 09:59:37 -03:00
oobabooga	2c2d453c8c	Revert "Use ExLlamaV2 (instead of the HF one) for EXL2 models for now" ... This reverts commit 0ef1b8f8b4.	2025-04-17 21:31:32 -07:00
oobabooga	0ef1b8f8b4	Use ExLlamaV2 (instead of the HF one) for EXL2 models for now ... It doesn't seem to have the "OverflowError" bug	2025-04-17 05:47:40 -07:00
oobabooga	8b8d39ec4e	Add ExLlamaV3 support (#6832)	2025-04-09 00:07:08 -03:00
oobabooga	5bcd2d7ad0	Add the top N-sigma sampler (#6796)	2025-03-14 16:45:11 -03:00
oobabooga	83c426e96b	Organize internals (#6646)	2025-01-10 18:04:32 -03:00
oobabooga	7157257c3f	Remove the AutoGPTQ loader (#6641)	2025-01-08 19:28:56 -03:00
oobabooga	c0f600c887	Add a --torch-compile flag for transformers	2025-01-05 05:47:00 -08:00
oobabooga	11af199aff	Add a "Static KV cache" option for transformers	2025-01-04 17:52:57 -08:00
oobabooga	3967520e71	Connect XTC, DRY, smoothing_factor, and dynatemp to ExLlamaV2 loader (non-HF)	2025-01-04 16:25:06 -08:00
oobabooga	39a5c9a49c	UI organization (#6618)	2024-12-29 11:16:17 -03:00
Diner Burger	addad3c63e	Allow more granular KV cache settings (#6561)	2024-12-17 17:43:48 -03:00
oobabooga	93c250b9b6	Add a UI element for enable_tp	2024-10-01 11:16:15 -07:00
Philipp Emanuel Weidmann	301375834e	Exclude Top Choices (XTC): A sampler that boosts creativity, breaks writing clichés, and inhibits non-verbatim repetition (#6335)	2024-09-27 22:50:12 -03:00
oobabooga	e6181e834a	Remove AutoAWQ as a standalone loader ... (it works better through transformers)	2024-07-23 15:31:17 -07:00
oobabooga	aa809e420e	Bump llama-cpp-python to 0.2.83, add back tensorcore wheels ... Also add back the progress bar patch	2024-07-22 18:05:11 -07:00
oobabooga	11bbf71aa5	Bump back llama-cpp-python (#6257)	2024-07-22 16:19:41 -03:00
oobabooga	0f53a736c1	Revert the llama-cpp-python update	2024-07-22 12:02:25 -07:00
oobabooga	a687f950ba	Remove the tensorcores llama.cpp wheels ... They are not faster than the default wheels anymore and they use a lot of space.	2024-07-22 11:54:35 -07:00
oobabooga	e436d69e2b	Add --no_xformers and --no_sdpa flags for ExllamaV2	2024-07-11 15:47:37 -07:00
GralchemOz	8a39f579d8	transformers: Add eager attention option to make Gemma-2 work properly (#6188)	2024-07-01 12:08:08 -03:00
oobabooga	4ea260098f	llama.cpp: add 4-bit/8-bit kv cache options	2024-06-29 09:10:33 -07:00
oobabooga	577a8cd3ee	Add TensorRT-LLM support (#5715)	2024-06-24 02:30:03 -03:00
oobabooga	536f8d58d4	Do not expose alpha_value to llama.cpp & rope_freq_base to transformers ... To avoid confusion	2024-06-23 22:09:24 -07:00
Forkoz	1d79aa67cf	Fix flash-attn UI parameter to actually store true. (#6076)	2024-06-13 00:34:54 -03:00
oobabooga	bd7cc4234d	Backend cleanup (#6025)	2024-05-21 13:32:02 -03:00
Philipp Emanuel Weidmann	852c943769	DRY: A modern repetition penalty that reliably prevents looping (#5677)	2024-05-19 23:53:47 -03:00
oobabooga	e61055253c	Bump llama-cpp-python to 0.2.69, add --flash-attn option	2024-05-03 04:31:22 -07:00
oobabooga	51fb766bea	Add back my llama-cpp-python wheels, bump to 0.2.65 (#5964)	2024-04-30 09:11:31 -03:00
oobabooga	9b623b8a78	Bump llama-cpp-python to 0.2.64, use official wheels (#5921)	2024-04-23 23:17:05 -03:00
oobabooga	d423021a48	Remove CTransformers support (#5807)	2024-04-04 20:23:58 -03:00
oobabooga	35da6b989d	Organize the parameters tab (#5767)	2024-03-28 16:45:03 -03:00
oobabooga	afb51bd5d6	Add StreamingLLM for llamacpp & llamacpp_HF (2nd attempt) (#5669)	2024-03-09 00:25:33 -03:00
oobabooga	2ec1d96c91	Add cache_4bit option for ExLlamaV2 (#5645)	2024-03-06 23:02:25 -03:00