github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-01 17:17:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

[Inference]Repalce Attention layer and MLP layer by shardformer to optimize the weight transpose operation，add fused_qkv and fused linear_add (#5340)

Browse Source 
* add fused qkv

* replace attn and mlp by shardformer

* fix bugs in mlp

* add docstrings

* fix test_inference_engine.py

* add optimize unbind

* add fused_addmm

* rm squeeze(1)

* refactor codes

* fix ci bugs

* rename ShardFormerLlamaMLP and ShardFormerLlamaAttention

* Removed the dependency on LlamaFlashAttention2

* rollback test_inference_engine.py
This commit is contained in:
yuehuayingxueluo
2024-02-01 15:49:39 +08:00
committed by GitHub
parent f8e456d202
commit 249644c23b
8 changed files with 510 additions and 341 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
examples/inference/run_benchmark.sh
View File

@@ -25,10 +25,20 @@ CUDA_VISIBLE_DEVICES_set_n_least_memory_usage 1
# benchmark llama2-7b one single GPU
for bsz in 16 32 64; do
python3 ${PY_SCRIPT} -m llama2-7b --tp_size 1 --pp_size 1 -b $bsz -s 512 --output_len 256 --mode ${mode} | tee logs/${mode}_${GPU}_${bsz}_256.txt
python3 ${PY_SCRIPT} -m llama2-7b --tp_size 1 --pp_size 1 -b $bsz -s 512 --output_len 256 --mode ${mode} | tee logs/${mode}_${GPU}_${bsz}_512_256.txt
done
for bsz in 16 32 64; do
python3 ${PY_SCRIPT} -m llama2-7b --tp_size 1 --pp_size 1 -b $bsz -s 1024 --output_len 256 --mode ${mode} | tee logs/${mode}_${GPU}_${bsz}_1024.txt
python3 ${PY_SCRIPT} -m llama2-7b --tp_size 1 --pp_size 1 -b $bsz -s 1024 --output_len 256 --mode ${mode} | tee logs/${mode}_${GPU}_${bsz}_1024_256.txt
done
for bsz in 16 32 64; do
python3 ${PY_SCRIPT} -m llama2-7b --tp_size 1 --pp_size 1 -b $bsz -s 256 --output_len 128 --mode ${mode} | tee logs/${mode}_${GPU}_${bsz}_256_128.txt
done
for bsz in 16 32 64; do
python3 ${PY_SCRIPT} -m llama2-7b --tp_size 1 --pp_size 1 -b $bsz -s 1024 --output_len 128 --mode ${mode} | tee logs/${mode}_${GPU}_${bsz}_1024_128.txt
done