Commit graph

Select branches

Hide pull requests

main

Mono Color

• 5b83493b1a
  Add ipex-llm npu option in setup.py (#11858) SONG Ge 2024-08-20 17:29:49 +0800
• ee6852c915
  Fix typo (#11862) Heyang Sun 2024-08-20 16:38:11 +0800
• 2946420e14
  add minicpmv 2.6 load_low_bit workaround (#11856) Yishuo Wang 2024-08-20 11:16:02 +0800
• 7380823f3f
  Update Llama2 multi-processes example (#11852) SONG Ge 2024-08-19 19:49:01 +0800
• 99b05ba1dc
  separate prefill into a process (#11787) Yang Wang 2024-08-19 02:53:36 -0700
• da3d7a3a53
  delete transformers version requirement (#11845) Jinhe 2024-08-19 17:53:02 +0800
• a0fbda5bc8
  add MiniCPM-Llama3-V-2_5 into all-in-one benchmark (#11849) Ruonan Wang 2024-08-19 02:51:16 -0700
• 9490781aec
  optimize phi3 memory usage again (#11848) Yishuo Wang 2024-08-19 17:26:59 +0800
• 3cd4e87168
  Support compress KV with quantize KV (#11812) Yina Chen 2024-08-19 10:32:32 +0300
• 6841a9ac8f
  fix load low bit com dtype (#11832) Zhao Changmin 2024-08-19 13:43:19 +0800
• cfc959defa
  Fixes regarding utf-8 in all-in-one benchmark (#11839) Yuwen Hu 2024-08-19 10:38:00 +0800
• 46a1cbfa64
  feat: add mixed_precision argument on ppl longbench evaluation (#11837) Chu,Youcheng 2024-08-19 10:00:44 +0800
• 580c94d0e2
  Remove gemma-2-9b-it 3k input from igpu-perf (#11834) Yuwen Hu 2024-08-17 13:10:05 +0800
• 9f17234f3b
  Add MiniCPM-V-2_6 to iGPU Perf (#11810) Jin, Qiao 2024-08-16 18:41:21 +0800
• 96796f95cb
  Update all-in-one benchmark prompts for continuation task & lookup update for minicpmv (#11827) Yuwen Hu 2024-08-16 17:16:35 +0800
• e966e85df8
  force lm_head optimization in any model if set environment variable (#11830) Yishuo Wang 2024-08-16 16:48:45 +0800
• 3b630fb9df
  updated ppl README (#11807) RyuKosei 2024-08-16 15:49:25 +0800
• e07a55665c
  Codegeex2 tokenization fix (#11831) Jinhe 2024-08-16 15:48:47 +0800
• a508b0a902
  added link to minicpm-v-2_6 example (#11829) Jinhe 2024-08-16 14:49:23 +0800
• adfbb9124a
  Reorganize MiniCPM-V-2_6 example & update others MiniCPM-V-2 exmaples (#11815) Jinhe 2024-08-16 14:48:56 +0800
• f463268e36
  fix: add run oneAPI instruction for the example of codeshell (#11828) Chu,Youcheng 2024-08-16 14:29:06 +0800
• 17a0beb21f
  optimize qwen2-audio again (#11825) Yishuo Wang 2024-08-16 11:11:35 +0800
• 6a8d07ddb4
  Update README.md (#11824) Jason Dai 2024-08-16 10:22:02 +0800
• 9e9086cc2a
  Update IPEX_LLM_PERFORMANCE_MODE (#11823) Yuwen Hu 2024-08-16 09:48:36 +0800
• 5a80fd2633
  Fix lightweight-serving no streaming resp on mtl (#11822) Wang, Jian4 2024-08-16 09:43:03 +0800
• e70ae0638e
  Fix vLLM not convert issues (#11817) Guancheng Fu 2024-08-15 19:04:05 +0800
• 750d4ad5dc
  fix minicpm-v-2 fp16 (#11819) Yishuo Wang 2024-08-15 18:34:40 +0800
• 6543321f04
  Remove 4k igpu perf on gemma-2-9b-it (#11820) Yuwen Hu 2024-08-15 18:06:19 +0800
• 28d1c972da
  add mixed_precision argument on ppl wikitext evaluation (#11813) Chu,Youcheng 2024-08-15 17:58:53 +0800
• 828ab16537
  fix phi3 and minicpmv cpu (#11818) Yishuo Wang 2024-08-15 17:43:29 +0800
• 4e178f0c5d
  rewrite minicpmv optimization (#11816) Yishuo Wang 2024-08-15 17:27:12 +0800
• 447c8ed324
  update transformers version for replit-code-v1-3b, `internlm2-chat-… (#11811) Ch1y0q 2024-08-15 16:40:48 +0800
• 2fbbb51e71
  transformers==4.37, yi & yuan2 & vicuna (#11805) Jinhe 2024-08-15 15:39:24 +0800
• f43da2d455
  deletion of specification of transformers version (#11808) Jinhe 2024-08-15 15:23:32 +0800
• 07b7f13982
  support and optimize qwen2-audio (#11809) Yishuo Wang 2024-08-15 14:59:04 +0800
• 3ac83f8396
  fix: delete ipex extension import in ppl wikitext evaluation (#11806) Chu,Youcheng 2024-08-15 13:40:01 +0800
• 016e840eed
  Fix performance tests (#11802) Yuwen Hu 2024-08-15 01:37:01 +0800
• e3c1dae619
  Fix Windows Unit Test (#11801) Shaojun Liu 2024-08-14 19:16:48 +0800
• 9a93808fc5
  fix and optimize minicpm v 2 (#11799) Yishuo Wang 2024-08-14 17:27:23 +0800
• d8d887edd2
  added minicpm-v-2_6 (#11794) Jinhe 2024-08-14 16:23:44 +0800
• 3d6cfa291d
  optimize minicpm v 2.5 (#11793) Yishuo Wang 2024-08-14 16:07:24 +0800
• 356281cb80
  Further all-in-one benchmark update continuation task (#11784) Yuwen Hu 2024-08-14 14:39:34 +0800
• 43cca3be27
  fix gemma2 runtime error caused by sliding window (#11788) Ruonan Wang 2024-08-14 05:43:33 +0300
• dbd14251dd
  Troubleshoot for sycl not found (#11774) Jinhe 2024-08-14 10:26:01 +0800
• 51bcac1229
  follow up on experimental support of fused decoder layer for llama2 (#11785) Yang Wang 2024-08-13 18:53:55 -0700
• cb79dcda93
  refactor llama convert to fix minicpm-v 2.5 optimization (#11783) Yishuo Wang 2024-08-14 09:29:57 +0800
• 7cd6ec9723
  MiniCPM-V support compresskv (#11779) Yina Chen 2024-08-13 14:03:40 +0300
• 3998de14f0
  Fix mistral forward_qkv in q4_0 (#11781) Qiyuan Gong 2024-08-13 16:48:19 +0800
• 70c828b87c
  deepspeed zero3 QLoRA finetuning (#11625) Heyang Sun 2024-08-13 16:15:29 +0800
• a184b120c9
  fix minicpm-v 2.5 (#11780) Yishuo Wang 2024-08-13 16:14:00 +0800
• ec184af243
  Add gemma-2-2b-it and gemma-2-9b-it to igpu nightly performance test (#11778) Yuwen Hu 2024-08-13 15:39:56 +0800
• a88c132e54
  Reduce Mistral softmax memory only in low memory mode (#11775) Qiyuan Gong 2024-08-13 14:50:54 +0800
• aa861df066
  use new fp32 softmax kernel (#11776) Yishuo Wang 2024-08-13 14:48:11 +0800
• 23d3acdc77
  Add experimental support of fused decoder layer for llama2 (#11768) binbin Deng 2024-08-13 14:41:36 +0800
• c28b3389e6
  Update npu multimodal example (#11773) Jin, Qiao 2024-08-13 14:14:59 +0800
• 81824ff8c9
  Fix stdout in all-in-one benchmark to utf-8 (#11772) Yuwen Hu 2024-08-13 10:51:08 +0800
• a1eb793f70
  optimize minicpm v 2_6 firs token perf (#11770) Yishuo Wang 2024-08-13 09:51:18 +0800
• 841dbcdf3a
  Fix compresskv with lookahead issue (#11767) Yina Chen 2024-08-12 13:53:55 +0300
• f97a77ea4e
  Update all-in-one benchmark for continuation task input preparation (#11760) Yuwen Hu 2024-08-12 17:49:45 +0800
• 1b05caba2b
  Set mistral fuse rope to false except fp6 & fp16 (#11765) Xu, Shuo 2024-08-12 17:25:07 +0800
• 8db34057b4
  optimize lookahead init time (#11769) Ruonan Wang 2024-08-12 12:19:12 +0300
• 05989ad0f9
  Update npu example and all in one benckmark (#11766) Jin, Qiao 2024-08-12 16:46:46 +0800
• 57d177738d
  optimize minicpm-v-2_6 repetition penalty (#11763) Yishuo Wang 2024-08-12 14:10:10 +0800
• fac4c01a6e
  Revert to use out-of-tree GPU driver (#11761) Shaojun Liu 2024-08-12 13:41:47 +0800
• 245dba0abc
  Fix lightweight-serving codegeex error (#11759) Wang, Jian4 2024-08-12 10:35:37 +0800
• 66fe2ee464
  initial support of IPEX_LLM_PERFORMANCE_MODE (#11754) Ruonan Wang 2024-08-09 14:04:09 +0300
• 4b9c57cc60
  Support compress kv with lookahead (#11752) Yina Chen 2024-08-09 12:39:57 +0300
• 93455aac09
  fix minicpm V 2.6 repeat output (#11753) Yishuo Wang 2024-08-09 17:39:24 +0800
• 7e917d6cfb
  fix gptq of llama (#11749) Ruonan Wang 2024-08-09 11:39:25 +0300
• dd46c141bd
  Phi3 support compresskv (#11733) Yina Chen 2024-08-09 10:43:43 +0300
• d8808cc2e3
  Mistral apply_rotary_pos_emb_no_cache_xpu use rope_theta from config (#11747) Qiyuan Gong 2024-08-09 10:35:51 +0800
• 044e486480
  Fix vLLM CPU /chat endpoint (#11748) Xiangyu Tian 2024-08-09 10:33:52 +0800
• 27b4b104ed
  Add qwen2-1.5b-instruct into igpu performance (#11735) Jinhe 2024-08-08 16:42:18 +0800
• 107f7aafd0
  enable inference mode for deepspeed tp serving (#11742) Shaojun Liu 2024-08-08 14:38:30 +0800
• 9e65cf00b3
  Add openai-whisper pytorch gpu (#11736) Zijie Li 2024-08-08 12:32:59 +0800
• 7e61fa1af7
  Revise GPU driver related guide in for Windows users (#11740) Yuwen Hu 2024-08-08 11:26:26 +0800
• d0c89fb715
  updated llama.cpp and ollama quickstart (#11732) Jinhe 2024-08-08 11:04:01 +0800
• 54cc9353db
  support and optimize minicpm-v-2_6 (#11738) Yishuo Wang 2024-08-07 18:21:16 +0800
• e956e71fc1
  fix conflict with quant kv (#11737) Yina Chen 2024-08-07 13:10:30 +0300
• 00a5574c8a
  Use merge_qkv to replace fused_qkv for llama2 (#11727) Ruonan Wang 2024-08-07 13:04:01 +0300
• d2abc9711b
  Fix MTL 4k input qwen2 compresskv error (#11734) Yina Chen 2024-08-07 11:21:57 +0300
• a71ae7c22b
  Support minicpm compresskv & modify default compresskv config & default enable compresskv on mtl 2.5k~4.5k (#11726) Yina Chen 2024-08-07 06:35:39 +0300
• c093f7d980
  fix phi3 (#11729) Yishuo Wang 2024-08-07 09:39:46 +0800
• e32d13d78c
  Remove Out of tree Driver from GPU driver installation document (#11728) Qiyuan Gong 2024-08-07 09:38:19 +0800
• e7f7141781
  Add benchmark util for transformers 4.42 (#11725) Zijie Li 2024-08-07 08:48:07 +0800
• 4676af2054
  add gemma2 example (#11724) Ch1y0q 2024-08-06 21:17:50 +0800
• 985213614b
  Removed no longer needed models for Arc nightly perf (#11722) SichengStevenLi 2024-08-06 16:12:00 +0800
• 929675aa6b
  support latest phi3 (#11721) Yishuo Wang 2024-08-06 15:52:55 +0800
• 11650b6f81
  upgrade glm-4v example transformers version (#11719) Jin, Qiao 2024-08-06 14:55:09 +0800
• bbdff6edeb
  optimize internvl2 4b performance (#11720) Yishuo Wang 2024-08-06 14:25:08 +0800
• f44b732aa8
  support internvl2-4b (#11718) Yishuo Wang 2024-08-06 13:36:32 +0800
• 7f241133da
  Add MiniCPM-Llama3-V-2_5 GPU example (#11693) Jin, Qiao 2024-08-06 10:22:41 +0800
• 808d9a7bae
  Add MiniCPM-V-2 GPU example (#11699) Jin, Qiao 2024-08-06 10:22:33 +0800
• 8fb36b9f4a
  add new benchmark_util.py (#11713) Zijie Li 2024-08-05 16:18:48 +0800
• 493cbd9a36
  Support lightweight-serving with internlm-xcomposer2-vl-7b multimodal input (#11703) Wang, Jian4 2024-08-05 09:36:04 +0800
• aa98ef96fe
  change mixed_precision to q6_k (#11706) Ruonan Wang 2024-08-02 10:55:16 +0300
• 1baa3efe0e
  Optimizations for Pipeline Parallel Serving (#11702) Xiangyu Tian 2024-08-02 12:06:59 +0800
• 8d1e0bd2f4
  add sdp causal support in llama (#11705) Yina Chen 2024-08-02 05:27:40 +0300
• 736a7ef72e
  add sdp_causal for mistral 4.36 (#11686) Ruonan Wang 2024-08-01 13:57:31 +0300
• 45c730ff39
  Chatglm support compresskv (#11690) Yina Chen 2024-08-01 13:20:20 +0300