[executorch][cuda] Optimize short-query INT4 matvec kernels #7903

Job	Run time
CI run decision / decide	31s
Get changed files / get-changed-files	4s
export-model-cuda-windows-artifact (mistralai, Voxtral-Mini-3B-2507, quantized-int4-weight-only) / linux-job	33m 10s
export-model-cuda-windows-artifact (nvidia, parakeet-tdt, non-quantized) / linux-job	23m 30s
export-model-cuda-windows-artifact (facebook, dinov2-small-imagenet1k-1-layer, non-quantized) / linux-job	15m 49s
export-model-cuda-windows-artifact (mistralai, Voxtral-Mini-3B-2507, non-quantized) / linux-job	32m 22s
export-model-cuda-windows-artifact (mistralai, Voxtral-Mini-4B-Realtime-2602, quantized-int4-tile... / linux-job	39m 29s
export-model-cuda-windows-artifact (nvidia, parakeet-tdt, quantized-int4-weight-only) / linux-job	23m 49s
test-model-cuda-windows-e2e (facebook, dinov2-small-imagenet1k-1-layer, non-quantized) / windows-job	33m 28s
test-model-cuda-windows-e2e (mistralai, Voxtral-Mini-3B-2507, non-quantized) / windows-job	33m 4s
test-model-cuda-windows-e2e (mistralai, Voxtral-Mini-4B-Realtime-2602, quantized-int4-tile-packed) / windows-job	32m 34s
test-model-cuda-windows-e2e (mistralai, Voxtral-Mini-3B-2507, quantized-int4-weight-only) / windows-job	33m 6s
test-model-cuda-windows-e2e (nvidia, parakeet-tdt, non-quantized) / windows-job	31m 17s
test-model-cuda-windows-e2e (nvidia, parakeet-tdt, quantized-int4-weight-only) / windows-job	31m 50s
	6h 4m 3s

Navigation Menu