3a2a7d88db
The accuracy results running on NPU Altlas A2 have changed, updating
reports for: All models (Qwen3-30B-A3B, Qwen2.5-VL-7B-Instruct,
Qwen3-8B-Base, DeepSeek-V2-Lite)
- [Workflow run][1]
[1]:
https://github.com/vllm-project/vllm-ascend/actions/runs/17459225764
- vLLM version: v0.10.1.1
- vLLM main:
2b30afa442
Signed-off-by: vllm-ascend-ci <vllm-ascend-ci@users.noreply.github.com>
Co-authored-by: vllm-ascend-ci <vllm-ascend-ci@users.noreply.github.com>
964 B
964 B
deepseek-ai/DeepSeek-V2-Lite
- vLLM Version: vLLM: 0.10.1.1 (1da94e6), vLLM Ascend Version: v0.10.1rc1 (7e16b4a)
- Software Environment: CANN: 8.2.RC1, PyTorch: 2.7.1, torch-npu: 2.7.1.dev20250724
- Hardware Environment: Atlas A2 Series
- Parallel mode: TP2
- Execution mode: ACLGraph
Command:
export MODEL_ARGS='pretrained=deepseek-ai/DeepSeek-V2-Lite,tensor_parallel_size=2,dtype=auto,trust_remote_code=True,max_model_len=4096,enforce_eager=True'
lm_eval --model vllm --model_args $MODEL_ARGS --tasks gsm8k \
--batch_size auto
| Task | Metric | Value | Stderr |
|---|---|---|---|
| gsm8k | exact_match,strict-match | ✅0.3813 | ± 0.0134 |
| gsm8k | exact_match,flexible-extract | ✅0.3836 | ± 0.0134 |