intel/xFasterTransformer

Issues

• Segmentation fault (core dumped) FIRST_TOKEN_WEIGHT_LOCATION=$1 NEXT_TOKEN_WEIGHT_LOCATION=$2 OMP_NUM_THREADS=$3 numactl -C $cpu_index -p $2 $BENCHMARK

  #473 opened 3 months ago by LittleNoob2333
  2

• Update Broken QR Code Link for WeChat on Wiki Page

  #472 opened 4 months ago by w1ida
  1

• oneDNN and XFT performance

  #471 opened 4 months ago by LittleNoob2333
  0

• [bug] HBM flat QUAD mode determination method is incorrect

  #446 opened 4 months ago by xuyizhan
  1

• about benchmark Illegal instruction

  #469 opened 4 months ago by LittleNoob2333
  1

• qwen1.5-32b long text input issue

  #411 opened 5 months ago by zhm-algo
  5

• [run_benchmark.sh] Few cores are running on HBM when batch-size >16 or 32

  #455 opened 5 months ago by hangfu-guo
  3

• [request]qwen1 not supported by vllm-xft

  #447 opened 5 months ago by zhm-algo
  3

• Crash when using CB mode with multi-rank

  #440 opened 5 months ago by a3213105
  0

• gcc8.2 编译报错

  #419 opened 5 months ago by bukejiyu
  3

• support vllm?

  #250 opened 6 months ago by leiwen83
  2

• llama-2-70b memory usage

  #403 opened 6 months ago by zhm-algo
  2

• Can we summarize the meanings of data type like bf16_fp16?, for example, what's activation data type and output data type, what's the computing instruction?

  #414 opened 6 months ago by heagoo
  1

• error quantization with AWQ + AutoGPTQ

  #312 opened 6 months ago by zhm-algo
  7

• chatglm3 6b error

  #335 opened 7 months ago by zhm-algo
  1

• performance issue for opt-1.3b with BS=1 BF16

  #339 opened 7 months ago by bin1guo
  1

• [output issue] found mistakes in llama-3-70b output by bf16_int4 during benchmark

  #413 opened 6 months ago by intelyoungway
  1

• BF16_INT4 model loading too slow

  #235 opened 6 months ago by intelyoungway
  1

• What's the meaning of bf16_int4 datatype?

  #395 opened 6 months ago by LeiZhou-97
  4

• llama-2-7B benchmarking error with chinese prompts

  #380 opened 6 months ago by qdym188
  1

• The current commit(d666741) of xFT make failed.

  #345 opened 7 months ago by xiuying1
  3

• Build Error: Failure to Download and Configure xdnn_lib

  #341 opened 7 months ago by Damonpkl
  1

• qwen2支持吗

  #300 opened 7 months ago by wzg-zhuo
  2

• [bug] Met some problems while following *step by step tutorial*

  #328 opened 7 months ago by lum1n0us
  2

• [bug] Need to install protobuf library when run benchmark in docker

  #282 opened 8 months ago by xuyizhan
  2

• 4 sockets qwen execute question

  #269 opened 8 months ago by Storm0921
  13

• xft will be blocked when MPI + QWEN14B + do_sample=true

  #204 opened 8 months ago by a3213105
  4

• Illegal instruction (core dumped)

  #247 opened 8 months ago by wswsmao
  5

• typo in benchmark script

  #255 opened 9 months ago by sssssux
  1

• torch==2.2.0 run error

  #233 opened 9 months ago by Zjq9409
  3

• Error install xfastertransformer on CentOS 7.6

  #244 opened 9 months ago by bin1guo
  1

• xft + QWEN14B + fp16 got unexpected outputs compared with bf16_fp16.

  #220 opened 9 months ago by a3213105
  2

• [BUG] PR #224 cause the wrong output of QWEN-14B

  #239 opened 9 months ago by a3213105
  0

• [bug] library of Intel level-zero not found

  #134 opened 9 months ago by intelyoungway
  3

• Qwen Segmentation Fault after logN PR merged.

  #229 opened 9 months ago by Duyi-Wang
  2

• xft + sample output result look bad

  #209 opened 9 months ago by Zjq9409
  5

• baichuan-7b run core dump

  #210 opened 10 months ago by Zjq9409
  1

• Use mpirun to run benchmark.py get error

  #153 opened 10 months ago by yangkunx
  4

• Illegal instruction (core dumped)

  #186 opened 10 months ago by Storm8878
  8

• QWEN14B will generate error output when multi queries with long input tokens.

  #195 opened 10 months ago by a3213105
  0

• [Model] QWen14B-Chat got wrong output when input tokens is too long

  #174 opened 10 months ago by a3213105
  2

• [Model] Convert HF Qwen model into FP16 data type

  #142 opened 10 months ago by changqi1
  1

• Qwen-14B-Chat转换问题

  #185 opened 10 months ago by Storm0921
  7

• encounter Qwen 72B stop id issue

  #188 opened 10 months ago by marvin-Yu
  0

• 这块CPU是否满足xFasterTF的最低要求？

  #182 opened 10 months ago by Storm0921
  3

• AMX_int8 not really be used when dtype="int8"

  #175 opened 10 months ago by yiangyang
  2

• core per numa calculation error

  #152 opened 10 months ago by zhm-algo
  3

• chatGLM2-6B crash while running 4 ranks with the datatype w8a8

  #155 opened 10 months ago by shanzhou2186
  0

• SHM reduceAdd performance issue on HBM with 2 sockets

  #154 opened a year ago by abenmao
  2

• [bug] Segmentation fault occurs at large batch sizes

  #140 opened a year ago by aurora327
  1