optimize qwen2 model on Gaudi #233
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
extra mark_step() was added on each decode layer Signed-off-by: Bob Zhu <[email protected]>
szutenberg
requested changes
Sep 12, 2024
There was a problem hiding this comment.
The change is similar to what we have in https://github.com/HabanaAI/vllm-fork/blob/habana_main/vllm/model_executor/models/llama.py#L323 -> LGTM
Signed-off-by: Bob Zhu <[email protected]>
szutenberg
approved these changes
Sep 13, 2024
|
@szutenberg , i just fixed a importing order issue on the PR which is identified by CI system. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add extra mark_step() on each decode layer to optimize the performance on Gaudi.
When by default, the performance with below command:
VLLM_SKIP_WARMUP=true python benchmark_throughput.py --model /local_dataset_2/pytorch/Qwen2-7B-Instruct/ --device hpu --seed 2024 --backend vllm --input-len 1000 --output-len 500 --num-prompts 200 --dtype bfloat16is:
Throughput: 1.34 requests/s, 2015.75 tokens/sAfter applying this patch, the performance boosts to:
Throughput: 2.67 requests/s, 4003.20 tokens/s