Skip to content

Actions: groxaxo/GPTQ-Pro

Actions

Kernel Format Diagnostics

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
43 workflow runs
43 workflow runs

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #43: Pull request #11 synchronize by groxaxo
11s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
11s
style(kernel): apply exact Ruff format to native-qweight tests Kernel Format Diagnostics #42: Commit 166a1fe pushed by groxaxo
11s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
11s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #41: Pull request #11 synchronize by groxaxo
16s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
16s
style(kernel): apply exact Ruff format after native-qweight pivot Kernel Format Diagnostics #40: Commit 6494e7c pushed by groxaxo
16s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
16s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #39: Pull request #11 synchronize by groxaxo
16s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
16s
docs(roadmap): record native-qweight runtime and compile gates Kernel Format Diagnostics #38: Commit d8ae680 pushed by groxaxo
14s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
14s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #37: Pull request #11 synchronize by groxaxo
12s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
12s
docs(readme): document native-qweight Ampere runtime Kernel Format Diagnostics #36: Commit e81ef38 pushed by groxaxo
13s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
13s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #35: Pull request #11 synchronize by groxaxo
11s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
11s
test(kernel): lock native-qweight runtime contract Kernel Format Diagnostics #34: Commit a399dae pushed by groxaxo
10s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
10s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #33: Pull request #11 synchronize by groxaxo
14s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
14s
perf(bench): benchmark native qweight without duplicate packing Kernel Format Diagnostics #32: Commit 250aaaa pushed by groxaxo
19s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
19s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #31: Pull request #11 synchronize by groxaxo
17s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
17s
test(cuda): validate native qweight dispatch paths Kernel Format Diagnostics #30: Commit 19898be pushed by groxaxo
13s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
13s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #29: Pull request #11 synchronize by groxaxo
19s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
19s
perf(runtime): remove duplicate packed-weight buffer Kernel Format Diagnostics #28: Commit 7d714b3 pushed by groxaxo
15s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
15s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #27: Pull request #11 synchronize by groxaxo
10s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
10s
perf(runtime): pass native qweight directly to CUDA Kernel Format Diagnostics #26: Commit 8446766 pushed by groxaxo
12s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
12s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #25: Pull request #11 synchronize by groxaxo
15s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
15s
perf(cuda): bind native int32 qweight without repacking Kernel Format Diagnostics #24: Commit 1f97a39 pushed by groxaxo
11s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
11s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #23: Pull request #11 synchronize by groxaxo
10s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
10s
perf(cuda): run all kernels directly from native qweight Kernel Format Diagnostics #22: Commit 33ec1e3 pushed by groxaxo
10s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
10s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #21: Pull request #11 synchronize by groxaxo
16s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
16s
perf(cuda): consume native GPTQ int32 words directly Kernel Format Diagnostics #20: Commit 39584ad pushed by groxaxo
11s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
11s
Optimize GPTQ-Pro kernels for Ampere decode and prefill Kernel Format Diagnostics #19: Pull request #11 synchronize by groxaxo
16s perf/ampere-kernel-pipeline
perf/ampere-kernel-pipeline
16s