Yishuo Wang
ad96f32ce0
optimize phi3 1st token performance ( #10981 )
2024-05-10 17:33:46 +08:00
Cengguang Zhang
cfed76b2ed
LLM: add long-context support for Qwen1.5-7B/Baichuan2-7B/Mistral-7B. ( #10937 )
...
* LLM: add split tensor support for baichuan2-7b and qwen1.5-7b.
* fix style.
* fix style.
* fix style.
* add support for mistral and fix condition threshold.
* fix style.
* fix comments.
2024-05-10 16:40:15 +08:00
Kai Huang
a6342cc068
Empty cache after phi first attention to support 4k input ( #10972 )
...
* empty cache
* fix style
2024-05-09 19:50:04 +08:00
Yishuo Wang
e753125880
use fp16_sdp when head_dim=96 ( #10976 )
2024-05-09 17:02:59 +08:00
Yishuo Wang
697ca79eca
use quantize kv and sdp in phi3-mini ( #10973 )
2024-05-09 15:16:18 +08:00
Wang, Jian4
3209d6b057
Fix spculative llama3 no stop error ( #10963 )
...
* fix normal
* add eos_tokens_id on sp and add list if
* update
* no none
2024-05-08 17:09:47 +08:00
Yishuo Wang
2ebec0395c
optimize phi-3-mini-128 ( #10959 )
2024-05-08 16:33:17 +08:00
Zhao Changmin
0d6e12036f
Disable fast_init_ in load_low_bit ( #10945 )
...
* fast_init_ disable
2024-05-08 10:46:19 +08:00
Yishuo Wang
c801c37bc6
optimize phi3 again: use quantize kv if possible ( #10953 )
2024-05-07 17:26:19 +08:00
Yishuo Wang
aa2fa9fde1
optimize phi3 again: use sdp if possible ( #10951 )
2024-05-07 15:53:08 +08:00
Qiyuan Gong
d7ca5d935b
Upgrade Peft version to 0.10.0 for LLM finetune ( #10886 )
...
* Upgrade Peft version to 0.10.0
* Upgrade Peft version in ARC unit test and HF-Peft example.
2024-05-07 15:09:14 +08:00
Wang, Jian4
191b184341
LLM: Optimize cohere model ( #10878 )
...
* use mlp and rms
* optimize kv_cache
* add fuse qkv
* add flash attention and fp16 sdp
* error fp8 sdp
* fix optimized
* fix style
* update
* add for pp
2024-05-07 10:19:50 +08:00
Guancheng Fu
49ab5a2b0e
Add embeddings ( #10931 )
2024-05-07 09:07:02 +08:00
Wang, Jian4
0e0bd309e2
LLM: Enable Speculative on Fastchat ( #10909 )
...
* init
* enable streamer
* update
* update
* remove deprecated
* update
* update
* add gpu example
2024-05-06 10:06:20 +08:00
Cengguang Zhang
75dbf240ec
LLM: update split tensor conditions. ( #10872 )
...
* LLM: update split tensor condition.
* add cond for split tensor.
* update priority of env.
* fix style.
* update env name.
2024-04-30 17:07:21 +08:00
Guancheng Fu
2c64754eb0
Add vLLM to ipex-llm serving image ( #10807 )
...
* add vllm
* done
* doc work
* fix done
* temp
* add docs
* format
* add start-fastchat-service.sh
* fix
2024-04-29 17:25:42 +08:00
Yishuo Wang
d884c62dc4
remove new_layout parameter ( #10906 )
2024-04-29 10:31:50 +08:00
Guancheng Fu
fbcd7bc737
Fix Loader issue with dtype fp16 ( #10907 )
2024-04-29 10:16:02 +08:00
Guancheng Fu
c9fac8c26b
Fix sdp logic ( #10896 )
...
* fix
* fix
2024-04-28 22:02:14 +08:00
Yina Chen
015d07a58f
Fix lookahead sample error & add update strategy ( #10894 )
...
* Fix sample error & add update strategy
* add mtl config
* fix style
* remove print
2024-04-28 17:21:00 +08:00
Cengguang Zhang
9752ffe979
LLM: update split qkv native sdp. ( #10895 )
...
* LLM: update split qkv native sdp.
* fix typo.
2024-04-26 18:47:35 +08:00
Guancheng Fu
990535b1cf
Add tensor parallel for vLLM ( #10879 )
...
* initial
* test initial tp
* initial sup
* fix format
* fix
* fix
2024-04-26 17:10:49 +08:00
Yishuo Wang
46ba962168
use new quantize kv ( #10888 )
2024-04-26 14:42:17 +08:00
Wang, Jian4
3e8ed54270
LLM: Fix bigdl_ipex_int8 warning ( #10890 )
2024-04-26 11:18:44 +08:00
Yina Chen
8811f268ff
Use new fp16 sdp in Qwen and modify the constraint ( #10882 )
2024-04-25 19:23:37 +08:00
Yang Wang
1ce8d7bcd9
Support the desc_act feature in GPTQ model ( #10851 )
...
* support act_order
* update versions
* fix style
* fix bug
* clean up
2024-04-24 10:17:13 -07:00
Yina Chen
dc27b3bc35
Use sdp when rest token seq_len > 1 in llama & mistral (for lookup & spec) ( #10790 )
...
* update sdp condition
* update
* fix
* update & test llama
* mistral
* fix style
* update
* fix style
* remove pvc constrain
* update ds on arc
* fix style
2024-04-24 17:24:01 +08:00
binbin Deng
c9feffff9a
LLM: support Qwen1.5-MoE-A2.7B-Chat pipeline parallel inference ( #10864 )
2024-04-24 16:02:27 +08:00
Yishuo Wang
2d210817ff
add phi3 optimization ( #10871 )
2024-04-24 15:17:40 +08:00
Cengguang Zhang
763413b7e1
LLM: support llama split tensor for long context in transformers>=4.36. ( #10844 )
...
* LLm: support llama split tensor for long context in transformers>=4.36.
* fix dtype.
* fix style.
* fix style.
* fix style.
* fix style.
* fix dtype.
* fix style.
2024-04-23 16:13:25 +08:00
ZehuaCao
92ea54b512
Fix speculative decoding bug ( #10855 )
2024-04-23 14:28:31 +08:00
Wang, Jian4
18c032652d
LLM: Add mixtral speculative CPU example ( #10830 )
...
* init mixtral sp example
* use different prompt_format
* update output
* update
2024-04-23 10:05:51 +08:00
Yishuo Wang
fe5a082b84
add phi-2 optimization ( #10843 )
2024-04-22 18:56:47 +08:00
Guancheng Fu
47bd5f504c
[vLLM]Remove vllm-v1, refactor v2 ( #10842 )
...
* remove vllm-v1
* fix format
2024-04-22 17:51:32 +08:00
Wang, Jian4
23c6a52fb0
LLM: Fix ipex torchscript=True error ( #10832 )
...
* remove
* update
* remove torchscript
2024-04-22 15:53:09 +08:00
Yina Chen
3daad242b8
Fix No module named 'transformers.cache_utils' with transformers < 4.36 ( #10835 )
...
* update sdp condition
* update
* fix
* fix 431 error
* revert sdp & style fix
* fix
* meet comments
2024-04-22 14:05:50 +08:00
Guancheng Fu
caf75beef8
Disable sdpa ( #10814 )
2024-04-19 17:33:18 +08:00
Yishuo Wang
57edf2033c
fix lookahead with transformers >= 4.36 ( #10808 )
2024-04-19 16:24:56 +08:00
Ovo233
1a885020ee
Updated importing of top_k_top_p_filtering for transformers>=4.39.0 ( #10794 )
...
* In transformers>=4.39.0, the top_k_top_p_filtering function has been deprecated and moved to the hugging face package trl. Thus, for versions >= 4.39.0, import this function from trl.
2024-04-19 15:34:39 +08:00
Yishuo Wang
08458b4f74
remove rms norm copy ( #10793 )
2024-04-19 13:57:48 +08:00
Ruonan Wang
754b0ffecf
Fix pvc llama ( #10798 )
...
* ifx
* update
2024-04-18 10:44:57 -07:00
Ruonan Wang
439c834ed3
LLM: add mixed precision for lm_head ( #10795 )
...
* add mixed_quantization
* meet code review
* update
* fix style
* meet review
2024-04-18 19:11:31 +08:00
Yina Chen
8796401b08
Support q4k in ipex-llm ( #10796 )
...
* support q4k
* update
2024-04-18 18:55:28 +08:00
Ruonan Wang
0e8aac19e3
add q6k precision in ipex-llm ( #10792 )
...
* add q6k
* add initial 16k
* update
* fix style
2024-04-18 16:52:09 +08:00
Wang, Jian4
14ca42a048
LLM：Fix moe indexs error on cpu ( #10791 )
2024-04-18 15:56:52 +08:00
Guancheng Fu
cbe7b5753f
Add vLLM[xpu] related code ( #10779 )
...
* Add ipex-llm side change
* add runable offline_inference
* refactor to call vllm2
* Verified async server
* add new v2 example
* add README
* fix
* change dir
* refactor readme.md
* add experimental
* fix
2024-04-18 15:29:20 +08:00
Wang, Jian4
209c3501e6
LLM: Optimize qwen1.5 moe model ( #10706 )
...
* update moe block
* fix style
* enable optmize MLP
* enabel kv_cache
* enable fuse rope
* enable fused qkv
* enable flash_attention
* error sdp quantize
* use old api
* use fuse
* use xetla
* fix python style
* update moe_blocks num
* fix output error
* add cpu sdpa
* update
* update
* update
2024-04-18 14:54:05 +08:00
Ziteng Zhang
ff040c8f01
LISA Finetuning Example ( #10743 )
...
* enabling xetla only supports qtype=SYM_INT4 or FP8E5
* LISA Finetuning Example on gpu
* update readme
* add licence
* Explain parameters of lisa & Move backend codes to src dir
* fix style
* fix style
* update readme
* support chatglm
* fix style
* fix style
* update readme
* fix
2024-04-18 13:48:10 +08:00
Yang Wang
952e517db9
use config rope_theta ( #10787 )
...
* use config rope_theta
* fix style
2024-04-17 20:39:11 -07:00
Guancheng Fu
31ea2f9a9f
Fix wrong output for Llama models on CPU ( #10742 )
2024-04-18 11:07:27 +08:00
