Skip to content
Self-hosted runner (benchmark)

Optimize Qwen2VL vision model by precomputing cos/sin embeds before ViT blocks #10095

Sign in to view logs

Optimize Qwen2VL vision model by precomputing cos/sin embeds before ViT blocks

Optimize Qwen2VL vision model by precomputing cos/sin embeds before ViT blocks #10095

Re-run triggered February 13, 2025 09:04
@zucchini-nlpzucchini-nlp
#35837
li-plus:fast-qwen2vl-vision-rope
Status Skipped
Total duration 3s
Artifacts –

benchmark.yml

on: pull_request
Matrix: Benchmark
Fit to window
Zoom out
Zoom in