binbin Deng
6ea1e71af0
Update PP inference benchmark script ( #11323 )
2024-06-17 09:59:36 +08:00
SONG Ge
be00380f1a
Fix pipeline parallel inference past_key_value error in Baichuan ( #11318 )
...
* fix past_key_value error
* add baichuan2 example
* fix style
* update doc
* add script link in doc
* fix import error
* update
2024-06-17 09:29:32 +08:00
Yina Chen
0af0102e61
Add quantization scale search switch ( #11326 )
...
* add scale_search switch
* remove llama3 instruct
* remove print
2024-06-14 18:46:52 +08:00
Ruonan Wang
8a3247ac71
support batch forward for q4_k, q6_k ( #11325 )
2024-06-14 18:25:50 +08:00
Yishuo Wang
e8dd8e97ef
fix chatglm lookahead on ARC ( #11320 )
2024-06-14 16:26:11 +08:00
Shaojun Liu
f5ef94046e
exclude dolly-v2-12b for arc perf test ( #11315 )
...
* test arc perf
* test
* test
* exclude dolly-v2-12b:2048
* revert changes
2024-06-14 15:35:56 +08:00
Xiangyu Tian
4359ab3172
LLM: Add /generate_stream endpoint for Pipeline-Parallel-FastAPI example ( #11187 )
...
Add /generate_stream and OpenAI-formatted endpoint for Pipeline-Parallel-FastAPI example
2024-06-14 15:15:32 +08:00
Jin Qiao
0e7a31a09c
ChatGLM Examples Restructure regarding Installation Steps ( #11285 )
...
* merge install step in glm examples
* fix section
* fix section
* fix tiktoken
2024-06-14 12:37:05 +08:00
Yishuo Wang
91965b5d05
add glm_sdpa back to fix chatglm-6b ( #11313 )
2024-06-14 10:31:43 +08:00
Yishuo Wang
7f65836cb9
fix chatglm2/3-32k/128k fp16 ( #11311 )
2024-06-14 09:58:07 +08:00
Xin Qiu
1b0c4c8cb8
use new rotary two in chatglm4 ( #11312 )
...
* use new rotary two in chatglm4
* rempve
2024-06-13 19:02:18 +08:00
Xin Qiu
f1410d6823
refactor chatglm4 ( #11301 )
...
* glm4
* remove useless code
* stype
* add rope_ratio
* update
* fix fp16
* fix style
2024-06-13 18:06:04 +08:00
Yishuo Wang
5e25766855
fix and optimize chatglm2-32k and chatglm3-128k ( #11306 )
2024-06-13 17:37:58 +08:00
binbin Deng
60cb1dac7c
Support PP for qwen1.5 ( #11300 )
2024-06-13 17:35:24 +08:00
binbin Deng
f97cce2642
Fix import error of ds autotp ( #11307 )
2024-06-13 16:22:52 +08:00
Jin Qiao
3682c6a979
add glm4 and qwen2 to igpu perf ( #11304 )
2024-06-13 16:16:35 +08:00
Yishuo Wang
a24666b8f3
fix chatglm3-6b-32k ( #11303 )
2024-06-13 16:01:34 +08:00
Yishuo Wang
01fe0fc1a2
refactor chatglm2/3 ( #11290 )
2024-06-13 12:22:58 +08:00
Guancheng Fu
57a023aadc
Fix vllm tp ( #11297 )
2024-06-13 10:47:48 +08:00
Ruonan Wang
986af21896
fix perf test( #11295 )
2024-06-13 10:35:48 +08:00
binbin Deng
220151e2a1
Refactor pipeline parallel multi-stage implementation ( #11286 )
2024-06-13 10:00:23 +08:00
Ruonan Wang
14b1e6b699
Fix gguf_q4k ( #11293 )
...
* udpate embedding parameter
* update benchmark
2024-06-12 20:43:08 +08:00
Yuwen Hu
8edcdeb0e7
Fix bug that torch.ops.torch_ipex.matmul_bias_out cannot work on Linux MTL for short input ( #11292 )
2024-06-12 19:12:57 +08:00
Wenjing Margaret Mao
b61f6e3ab1
Add update_parent_folder for nightly_perf_test ( #11287 )
...
* add update_parent_folder and change the workflow file
* add update_parent_folder and change the workflow file
* move to pr mode and comment the test
* use one model per comfig
* revert
---------
Co-authored-by: Yishuo Wang <yishuo.wang@intel.com>
2024-06-12 17:58:13 +08:00
Xin Qiu
592f7aa61e
Refine glm1-4 sdp ( #11276 )
...
* chatglm
* update
* update
* change chatglm
* update sdpa
* update
* fix style
* fix
* fix glm
* update glm2-32k
* update glm2-32k
* fix cpu
* update
* change lower_bound
2024-06-12 17:11:56 +08:00
Yuwen Hu
cffb932f05
Expose timeout for streamer for fastchat worker ( #11288 )
...
* Expose timeout for stremer for fastchat worker
* Change to read from env variables
2024-06-12 17:02:40 +08:00
ivy-lv11
e7a4e2296f
Add Stable Diffusion examples on GPU and CPU ( #11166 )
...
* add sdxl and lcm-lora
* readme
* modify
* add cpu
* add license
* modify
* add file
2024-06-12 16:33:25 +08:00
Jin Qiao
f224e98297
Add GLM-4 CPU example ( #11223 )
...
* Add GLM-4 example
* add tiktoken dependency
* fix
* fix
2024-06-12 15:30:51 +08:00
Zijie Li
40fc8704c4
Add GPU example for GLM-4 ( #11267 )
...
* Add GPU example for GLM-4
* Update streamchat.py
* Fix pretrianed arguments
Fix pretrained arguments in generate and streamchat.py
* Update Readme
Update install tiktoken required for GLM-4
* Update comments in generate.py
2024-06-12 14:29:50 +08:00
Qiyuan Gong
0d9cc9c106
Remove duplicate check for ipex ( #11281 )
...
* Replacing builtin.import is causing lots of unpredicted problems. Remove this function.
2024-06-12 13:52:02 +08:00
Yishuo Wang
10e480ee96
refactor internlm and internlm2 ( #11274 )
2024-06-11 14:19:19 +08:00
Yuwen Hu
fac49f15e3
Remove manual importing ipex in all-in-one benchmark ( #11272 )
2024-06-11 09:32:13 +08:00
Wenjing Margaret Mao
70b17c87be
Merge multiple batches ( #11264 )
...
* add merge steps
* move to pr mode
* remove build + add merge.py
* add tohtml and change cp
* change test_batch folder path
* change merge_temp path
* change to html folder
* revert
* change place
* revert 437
* revert space
---------
Co-authored-by: Yishuo Wang <yishuo.wang@intel.com>
2024-06-07 18:38:45 +08:00
Xiangyu Tian
4b07712fd8
LLM: Fix vLLM CPU model convert mismatch ( #11254 )
...
Fix vLLM CPU model convert mismatch.
2024-06-07 15:54:34 +08:00
Yishuo Wang
42fab480ea
support stablm2 12b ( #11265 )
2024-06-07 15:46:00 +08:00
Xin Qiu
dbc3c2d72d
glm4 sdp ( #11253 )
...
* glm4 sdp
* fix style
* update comment
2024-06-07 15:42:23 +08:00
Xin Qiu
151fcf37bb
check devie name in use_flash_attention ( #11263 )
2024-06-07 15:07:47 +08:00
Yishuo Wang
2623944604
qwen2 sdpa small fix ( #11261 )
2024-06-07 14:42:18 +08:00
Yishuo Wang
ea0d03fd28
Refactor baichuan1 7B and 13B ( #11258 )
2024-06-07 14:29:20 +08:00
Qiyuan Gong
1aa9c9597a
Avoid duplicate import in IPEX auto importer ( #11227 )
...
* Add custom import to avoid ipex duplicate importing
* Add scope limitation
2024-06-07 14:08:00 +08:00
Wang, Jian4
6f2684e5c9
Update pp llama.py to save memory ( #11233 )
2024-06-07 13:18:16 +08:00
Yishuo Wang
ef8e9b2ecd
Refactor qwen2 moe ( #11244 )
2024-06-07 13:14:54 +08:00
Zijie Li
7b753dc8ca
Update sample output for HF Qwen2 GPU and CPU ( #11257 )
2024-06-07 11:36:22 +08:00
Zhao Changmin
b7948671de
[WIP] Add look up table in 1st token stage ( #11193 )
...
* lookuptb
2024-06-07 10:51:05 +08:00
Yuwen Hu
8c36b5bdde
Add qwen2 example ( #11252 )
...
* Add GPU example for Qwen2
* Update comments in README
* Update README for Qwen2 GPU example
* Add CPU example for Qwen2
Sample Output under README pending
* Update generate.py and README for CPU Qwen2
* Update GPU example for Qwen2
* Small update
* Small fix
* Add Qwen2 table
* Update README for Qwen2 CPU and GPU
Update sample output under README
---------
Co-authored-by: Zijie Li <michael20001122@gmail.com>
2024-06-07 10:29:33 +08:00
Shaojun Liu
85df5e7699
fix nightly perf test ( #11251 )
2024-06-07 09:33:14 +08:00
Xin Qiu
2f809116e2
optimize Chatglm4 ( #11239 )
...
* chatglm4
* update
* update
* add rms norm
* chatglm4
2024-06-06 18:25:20 +08:00
hxsz1997
b6234eb4e2
Add task in allinone ( #11226 )
...
* add task
* update prompt
* modify typos
* add more cases in summarize
* Make the summarize & QA prompt preprocessing as a util function
2024-06-06 17:22:40 +08:00
Yishuo Wang
2e4ccd541c
fix qwen2 cpu ( #11240 )
2024-06-06 16:24:19 +08:00
Yishuo Wang
e738ec38f4
disable quantize kv in specific qwen model ( #11238 )
2024-06-06 14:08:39 +08:00
Yishuo Wang
c4e5806e01
add latest optimization in starcoder2 ( #11236 )
2024-06-06 14:02:17 +08:00
Yishuo Wang
ba27e750b1
refactor yuan2 ( #11235 )
2024-06-06 13:17:54 +08:00
Shaojun Liu
6be24fdd28
OSPDT: add tpp licenses ( #11165 )
...
* add tpp licenses
* add licenses
* add licenses
* delete mitchellh-mapstructure license
* delete stb-image public domain license
* add README.md
* remove core-xe related licenses
2024-06-06 10:59:06 +08:00
Guoqiong Song
09c6780d0c
phi-2 transformers 4.37 ( #11161 )
...
* phi-2 transformers 4.37
2024-06-05 13:36:41 -07:00
Guoqiong Song
f6d5c6af78
fix issue 1407 ( #11171 )
2024-06-05 13:35:57 -07:00
Zijie Li
bfa1367149
Add CPU and GPU example for MiniCPM ( #11202 )
...
* Change installation address
Change former address: "https://docs.conda.io/en/latest/miniconda.html# " to new address: "https://conda-forge.org/download/ " for 63 occurrences under python\llm\example
* Change Prompt
Change "Anaconda Prompt" to "Miniforge Prompt" for 1 occurrence
* Create and update model minicpm
* Update model minicpm
Update model minicpm under GPU/PyTorch-Models
* Update readme and generate.py
change "prompt = tokenizer.apply_chat_template(chat, tokenize=False, add_generation_prompt=False)" and delete "pip install transformers==4.37.0
"
* Update comments for minicpm GPU
Update comments for generate.py at minicpm GPU
* Add CPU example for MiniCPM
* Update minicpm README for CPU
* Update README for MiniCPM and Llama3
* Update Readme for Llama3 CPU Pytorch
* Update and fix comments for MiniCPM
2024-06-05 18:09:53 +08:00
Yuwen Hu
af96579c76
Update installation guide for pipeline parallel inference ( #11224 )
...
* Update installation guide for pipeline parallel inference
* Small fix
* further fix
* Small fix
* Small fix
* Update based on comments
* Small fix
* Small fix
* Small fix
2024-06-05 17:54:29 +08:00
Yina Chen
ed67435491
Support Fp6 k in ipex-llm ( #11222 )
...
* support fp6_k
* support fp6_k
* remove
* fix style
2024-06-05 17:34:36 +08:00
binbin Deng
a6674f5bce
Fix should_use_fuse_rope error of Qwen1.5-MoE-A2.7B-Chat ( #11216 )
2024-06-05 15:56:10 +08:00
Wenjing Margaret Mao
231b968aba
Modify the check_results.py to support batch 2&4 ( #11133 )
...
* add batch 2&4 and exclude to perf_test
* modify the perf-test&437 yaml
* modify llm_performance_test.yml
* remove batch 4
* modify check_results.py to support batch 2&4
* change the batch_size format
* remove genxir
* add str(batch_size)
* change actual_test_casese in check_results file to support batch_size
* change html highlight
* less models to test html and html_path
* delete the moe model
* split batch html
* split
* use installing from pypi
* use installing from pypi - batch2
* revert cpp
* revert cpp
* merge two jobs into one, test batch_size in one job
* merge two jobs into one, test batch_size in one job
* change file directory in workflow
* try catch deal with odd file without batch_size
* modify pandas version
* change the dir
* organize the code
* organize the code
* remove Qwen-MOE
* modify based on feedback
* modify based on feedback
* modify based on second round of feedback
* modify based on second round of feedback + change run-arc.sh mode
* modify based on second round of feedback + revert config
* modify based on second round of feedback + revert config
* modify based on second round of feedback + remove comments
* modify based on second round of feedback + remove comments
* modify based on second round of feedback + revert arc-perf-test
* modify based on third round of feedback
* change error type
* change error type
* modify check_results.html
* split batch into two folders
* add all models
* move csv_name
* revert pr test
* revert pr test
---------
Co-authored-by: Yishuo Wang <yishuo.wang@intel.com>
2024-06-05 15:04:55 +08:00
Xin Qiu
566691c5a3
quantized attention forward for minicpm ( #11200 )
...
* quantized minicpm
* fix style check
2024-06-05 09:15:25 +08:00
Jiao Wang
bb83bc23fd
Fix Starcoder issue on CPU on transformers 4.36+ ( #11190 )
...
* fix starcoder for sdpa
* update
* style
2024-06-04 10:05:40 -07:00
Kai Huang
f93664147c
Update config.yaml ( #11208 )
...
* update config.yaml
* fix
* minor
* style
2024-06-04 19:58:18 +08:00
Xiangyu Tian
ac3d53ff5d
LLM: Fix vLLM CPU version error ( #11206 )
...
Fix vLLM CPU version error
2024-06-04 19:10:23 +08:00
Ruonan Wang
1dde204775
update q6k ( #11205 )
2024-06-04 17:14:33 +08:00
Qiyuan Gong
ce3f08b25a
Fix IPEX auto importer ( #11192 )
...
* Fix ipex auto importer with Python builtins.
* Raise errors if the user imports ipex manually before importing ipex_llm. Do nothing if they import ipex after importing ipex_llm.
* Remove import ipex in examples.
2024-06-04 16:57:18 +08:00
Yina Chen
711fa0199e
Fix fp6k phi3 ppl core dump ( #11204 )
2024-06-04 16:44:27 +08:00
Xiangyu Tian
f02f097002
Fix vLLM verion in CPU/vLLM-Serving example README ( #11201 )
2024-06-04 15:56:55 +08:00
Yishuo Wang
6454655dcc
use sdp in baichuan2 13b ( #11198 )
2024-06-04 15:39:00 +08:00
Yishuo Wang
d90cd977d0
refactor stablelm ( #11195 )
2024-06-04 13:14:43 +08:00
Zijie Li
a644e9409b
Miniconda/Anaconda -> Miniforge update in examples ( #11194 )
...
* Change installation address
Change former address: "https://docs.conda.io/en/latest/miniconda.html# " to new address: "https://conda-forge.org/download/ " for 63 occurrences under python\llm\example
* Change Prompt
Change "Anaconda Prompt" to "Miniforge Prompt" for 1 occurrence
2024-06-04 10:14:02 +08:00
Xin Qiu
5f13700c9f
optimize Minicpm ( #11189 )
...
* minicpm optimize
* update
2024-06-03 18:28:29 +08:00
Qiyuan Gong
15a6205790
Fix LoRA tokenizer for Llama and chatglm ( #11186 )
...
* Set pad_token to eos_token if it's None. Otherwise, use model config.
2024-06-03 15:35:38 +08:00
Cengguang Zhang
3eb13ccd8c
LLM: fix input length condition in deepspeed all-in-one benchmark. ( #11185 )
2024-06-03 10:05:43 +08:00
Shaojun Liu
401013a630
Remove chatglm_C Module to Eliminate LGPL Dependency ( #11178 )
...
* remove chatglm_C.**.pyd to solve ngsolve weak copyright vunl
* fix style check error
* remove chatglm native int4 from langchain
2024-05-31 17:03:11 +08:00
Ruonan Wang
50b5f4476f
update q4k convert ( #11179 )
2024-05-31 11:36:53 +08:00
Wang, Jian4
c0f1be6aea
Fix pp logic ( #11175 )
...
* only send no none batch and rank1-n sending first
* always send first
2024-05-30 16:40:59 +08:00
ZehuaCao
4127b99ed6
Fix null pointer dereferences error. ( #11125 )
...
* delete unused function on tgi_server
* update
* update
* fix style
2024-05-30 16:16:10 +08:00
Guancheng Fu
50ee004ac7
Fix vllm condition ( #11169 )
...
* add use-vllm
* done
* fix style
* fix done
2024-05-30 15:23:17 +08:00
Jin Qiao
dcbf4d3d0a
Add phi-3-vision example ( #11156 )
...
* Add phi-3-vision example (HF-Automodels)
* fix
* fix
* fix
* Add phi-3-vision CPU example (HF-Automodels)
* add in readme
* fix
* fix
* fix
* fix
* use fp8 for gpu example
* remove eval
2024-05-30 10:02:47 +08:00
Jiao Wang
93146b9433
Reconstruct Speculative Decoding example directory ( #11136 )
...
* update
* update
* update
2024-05-29 13:15:27 -07:00
Xiangyu Tian
2299698b45
Refine Pipeline Parallel FastAPI example ( #11168 )
2024-05-29 17:16:50 +08:00
Ruonan Wang
9bfbf78bf4
update api usage of xe_batch & fp16 ( #11164 )
...
* update api usage
* update setup.py
2024-05-29 15:15:14 +08:00
Yina Chen
e29e2f1c78
Support new fp8 e4m3 ( #11158 )
2024-05-29 14:27:14 +08:00
Wang, Jian4
8e25de1126
LLM: Add codegeex2 example ( #11143 )
...
* add codegeex example
* update
* update cpu
* add GPU
* add gpu
* update readme
2024-05-29 10:00:26 +08:00
ZehuaCao
751e1a4e29
Fix concurrent issue in autoTP streming. ( #11150 )
...
* add benchmark test
* update
2024-05-29 08:22:38 +08:00
Yishuo Wang
bc5008f0d5
disable sdp_causal in phi-3 to fix overflow ( #11157 )
2024-05-28 17:25:53 +08:00
SONG Ge
33852bd23e
Refactor pipeline parallel device config ( #11149 )
...
* refactor pipeline parallel device config
* meet comments
* update example
* add warnings and update code doc
2024-05-28 16:52:46 +08:00
hxsz1997
62b2d8af6b
Add lookahead in all-in-one ( #11142 )
...
* add lookahead in allinone
* delete save to csv in run_transformer_int4_gpu
* change lookup to lookahead
* fix the error of add model.peak_memory
* Set transformer_int4_gpu as the default option
* add comment of transformer_int4_fp16_lookahead_gpu
2024-05-28 15:39:58 +08:00
Xiangyu Tian
b44cf405e2
Refine Pipeline-Parallel-Fastapi example README ( #11155 )
2024-05-28 15:18:21 +08:00
Yishuo Wang
d307622797
fix first token sdp with batch ( #11153 )
2024-05-28 15:03:06 +08:00
Yina Chen
3464440839
fix qwen import error ( #11154 )
2024-05-28 14:50:12 +08:00
Jin Qiao
25b6402315
Add Windows GPU unit test ( #11050 )
...
* Add Windows GPU UT
* temporarily remove ut on arc
* retry
* retry
* retry
* fix
* retry
* retry
* fix
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* fix
* retry
* retry
* retry
* retry
* retry
* retry
* merge into single workflow
* retry inference test
* retry
* retrigger
* try to fix inference test
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* retry
* check lower_bound
* retry
* retry
* try example test
* try fix example test
* retry
* fix
* seperate function into shell script
* remove cygpath
* try remove all cygpath
* retry
* retry
* Revert "try remove all cygpath"
This reverts commit 7ceeff3e48f08429062ecef548c1a3ad3488756f.
* Revert "retry"
This reverts commit 40ea2457843bff6991b8db24316cde5de1d35418.
* Revert "retry"
This reverts commit 817d0db3e5aec3bd449d3deaf4fb01d3ecfdc8a3.
* enable ut
* fix
* retrigger
* retrigger
* update download url
* fix
* fix
* retry
* add comment
* fix
2024-05-28 13:29:47 +08:00
Yina Chen
b6b70d1ba0
Divide core-xe packages ( #11131 )
...
* temp
* add batch
* fix style
* update package name
* fix style
* add workflow
* use temp version to run uts
* trigger performance test
* trigger win igpu perf
* revert workflow & setup
2024-05-28 12:00:18 +08:00
binbin Deng
c9168b85b7
Fix error during merging adapter ( #11145 )
2024-05-27 19:41:42 +08:00
Guancheng Fu
daf7b1cd56
[Docker] Fix image using two cards error ( #11144 )
...
* fix all
* done
2024-05-27 16:20:13 +08:00
Xiangyu Tian
5c8ccf0ba9
LLM: Add Pipeline-Parallel-FastAPI example ( #10917 )
...
Add multi-stage Pipeline-Parallel-FastAPI example
---------
Co-authored-by: hzjane <a1015616934@qq.com>
2024-05-27 14:46:29 +08:00
Ruonan Wang
d550af957a
fix security issue of eagle ( #11140 )
...
* fix security issue of eagle
* small fix
2024-05-27 10:15:28 +08:00
binbin Deng
367de141f2
Fix mixtral-8x7b with transformers=4.37.0 ( #11132 )
2024-05-27 09:50:54 +08:00
Jean Yu
ab476c7fe2
Eagle Speculative Sampling examples ( #11104 )
...
* Eagle Speculative Sampling examples
* rm multi-gpu and ray content
* updated README to include Arc A770
2024-05-24 11:13:43 -07:00
