Skip to content
Test CUDA Builds

SDPA decode perf improvements for qwen-3.5-35B-A3B #12572

Sign in to view logs

SDPA decode perf improvements for qwen-3.5-35B-A3B

SDPA decode perf improvements for qwen-3.5-35B-A3B #12572

Run time

Learn about OS pricing on GitHub Actions
Job Run time
unittest-cuda / linux-job
40m 17s
test-models-cuda / linux-job
43m 50s
test-executorch-cuda-build-13.0 / linux-job
16m 6s
test-executorch-cuda-build-12.9 / linux-job
21m 35s
test-executorch-cuda-build-12.6 / linux-job
16m 25s
test-executorch-cuda-build-12.8 / linux-job
17m 35s
export-model-cuda-artifact (openai, whisper-small, non-quantized) / linux-job
21m 4s
export-model-cuda-artifact (openai, whisper-large-v3-turbo, quantized-int4-weight-only) / linux-job
22m 21s
export-model-cuda-artifact (nvidia, parakeet-tdt, quantized-int4-tile-packed) / linux-job
27m 37s
export-model-cuda-artifact (SocialLocalMobile, Qwen3.5-35B-A3B-HQQ-INT4, quantized-int4-tile-packed) / linux-job
1h 4m 36s
export-model-cuda-artifact (openai, whisper-large-v3-turbo, quantized-int4-tile-packed) / linux-job
22m 29s
export-model-cuda-artifact (mistralai, Voxtral-Mini-3B-2507, quantized-int4-weight-only) / linux-job
31m 0s
export-model-cuda-artifact (Qwen, Qwen3-0.6B, quantized-int4-tile-packed) / linux-job
22m 1s
export-model-cuda-artifact (google, gemma-3-4b-it, quantized-int4-tile-packed) / linux-job
32m 53s
export-model-cuda-artifact (facebook, dinov2-small-imagenet1k-1-layer, non-quantized) / linux-job
18m 3s
export-model-cuda-artifact (nvidia, diar_streaming_sortformer_4spk-v2, non-quantized) / linux-job
32m 52s
export-model-cuda-artifact (google, gemma-3-4b-it, non-quantized) / linux-job
34m 8s
export-model-cuda-artifact (nvidia, parakeet-tdt, non-quantized) / linux-job
26m 2s
export-model-cuda-artifact (mistralai, Voxtral-Mini-3B-2507, non-quantized) / linux-job
30m 49s
export-model-cuda-artifact (mistralai, Voxtral-Mini-3B-2507, quantized-int4-tile-packed) / linux-job
33m 5s
export-model-cuda-artifact (Qwen, Qwen3-0.6B, non-quantized) / linux-job
22m 1s
export-model-cuda-artifact (mistralai, Voxtral-Mini-4B-Realtime-2602, quantized-int4-tile-packed) / linux-job
28m 34s
check-all-cuda-builds
2s
test-cuda-pybind (gemma3-4b, --quantize, google-gemma-3-4b-it-cuda-quantized-int4-tile-packed) / linux-job
21m 21s
test-cuda-pybind (qwen3-0.6b, --quantize, Qwen-Qwen3-0.6B-cuda-quantized-int4-tile-packed) / linux-job
19m 41s
test-cuda-pybind (qwen3-0.6b, Qwen-Qwen3-0.6B-cuda-non-quantized) / linux-job
19m 49s
test-model-cuda-e2e (mistralai, Voxtral-Mini-3B-2507, quantized-int4-tile-packed) / linux-job
14m 6s
test-model-cuda-e2e (mistralai, Voxtral-Mini-4B-Realtime-2602, quantized-int4-tile-packed) / linux-job
29m 10s
test-model-cuda-e2e (mistralai, Voxtral-Mini-3B-2507, quantized-int4-weight-only) / linux-job
15m 25s
test-model-cuda-e2e (google, gemma-3-4b-it, quantized-int4-tile-packed) / linux-job
14m 21s
test-model-cuda-e2e (openai, whisper-small, non-quantized) / linux-job
27m 36s
test-model-cuda-e2e (google, gemma-3-4b-it, non-quantized) / linux-job
15m 49s
test-model-cuda-e2e (nvidia, diar_streaming_sortformer_4spk-v2, non-quantized) / linux-job
13m 58s
test-model-cuda-e2e (facebook, dinov2-small-imagenet1k-1-layer, non-quantized) / linux-job
14m 10s
test-model-cuda-e2e (openai, whisper-large-v3-turbo, quantized-int4-tile-packed) / linux-job
28m 34s
test-model-cuda-e2e (SocialLocalMobile, Qwen3.5-35B-A3B-HQQ-INT4, quantized-int4-tile-packed) / linux-job
24m 18s
test-model-cuda-e2e (openai, whisper-large-v3-turbo, quantized-int4-weight-only) / linux-job
29m 9s
test-model-cuda-e2e (nvidia, parakeet-tdt, non-quantized) / linux-job
14m 44s
test-model-cuda-e2e (mistralai, Voxtral-Mini-3B-2507, non-quantized) / linux-job
15m 30s
test-model-cuda-e2e (nvidia, parakeet-tdt, quantized-int4-tile-packed) / linux-job
14m 12s
15h 57m 18s