vLLM
/
vllm-project.github.io
mirror of https://github.com/vllm-project/vllm-project.github.io.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #8 from tjtanaa/2024-10-23-vllm-serving-amd-amend-tj 

[FIX] Amend benchmark command and model data type
This commit is contained in:
Simon Mo 2024-10-30 09:28:57 -07:00 committed by GitHub
parent dc147caa3f 9769c02a65
commit cf68f9893b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
_posts/2024-10-23-vllm-serving-amd.md
View File

@ -14,7 +14,7 @@ author: "Guest Post by Embedded LLM and Hot Aisle Inc."
   
<img src="/assets/figures/vllm-serving-amd/405b2.png" width="35%">
</picture><br>
vLLM vs. TGI performance comparison for Llama 3.1 405B on 8 x MI300X (FP16, 32 QPS).
vLLM vs. TGI performance comparison for Llama 3.1 405B on 8 x MI300X (BF16, 32 QPS).
</p>
<p align="center">
@ -24,7 +24,7 @@ vLLM vs. TGI performance comparison for Llama 3.1 405B on 8 x MI300X (FP16, 32 Q
&nbsp; &nbsp;
<img src="/assets/figures/vllm-serving-amd/70b2.png" width="35%">
</picture><br>
vLLM vs. TGI performance comparison for Llama 3.1 70B on 8 x MI300X (FP16, 32 QPS).
vLLM vs. TGI performance comparison for Llama 3.1 70B on 8 x MI300X (BF16, 32 QPS).
</p>
### Introduction
@ -49,7 +49,7 @@ Even in the default configuration, vLLM shows superior performance compared to T
<picture>
<img src="/assets/figures/vllm-serving-amd/introduction/Mean TTFT (ms).png" width="70%">
</picture><br>
vLLM vs. TGI performance for Llama 3.1 405B on 8 x MI300X (FP16, QPS 16, 32, 1000; see Appendix for commands).
vLLM vs. TGI performance for Llama 3.1 405B on 8 x MI300X (BF16, QPS 16, 32, 1000; see Appendix for commands).
</p>
### How to run vLLM with Optimal Performance
@ -335,5 +335,5 @@ We have built the ROCm compatible vLLM docker from Dockerfile.rocm found in the
| ------------- | ------------- |
| vLLM Default Configuration | `VLLM_RPC_TIMEOUT=30000 VLLM_USE_TRITON_FLASH_ATTN=0 vllm serve Llama-3.1-405B-Instruct -tp 8 --max-num-seqs 1024 --max-num-batched-tokens 1024 ` |
| TGI Default Configuration | `ROCM_USE_FLASH_ATTN_V2_TRITON=false TRUST_REMOTE_CODE=true text-generation-launcher --num-shard 8 --sharded true --max-concurrent-requests 1024 --model-id Llama-3.1-405B-Instruct` |
| vLLM (This Guide) | `VLLM_RPC_TIMEOUT=30000 VLLM_USE_TRITON_FLASH_ATTN=0 vllm serve Llama-3.1-405B-Instruct-FP8 -tp 8 --max-seq-len-to-capture 16384 --enable-chunked-prefill=False --num-scheduler-step 15 --max-num-seqs 1024 ` |
| vLLM (This Guide) | `VLLM_RPC_TIMEOUT=30000 VLLM_USE_TRITON_FLASH_ATTN=0 vllm serve Llama-3.1-405B-Instruct -tp 8 --max-seq-len-to-capture 16384 --enable-chunked-prefill=False --num-scheduler-step 15 --max-num-seqs 1024 ` |
| TGI (This Guide) | `ROCM_USE_FLASH_ATTN_V2_TRITON=false TRUST_REMOTE_CODE=true text-generation-launcher --num-shard 8 --sharded true --max-concurrent-requests 1024 --max-total-tokens 131072 --max-input-tokens 131000 --model-id Llama-3.1-405B-Instruct` |