Skip to content

Actions: gittensor-ai-lab/sparkinfer

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
2,500+ workflow runs
2,500+ workflow runs

Filter by Workflow

⌥ + click/return to exclude

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
perf(muse): keep four packed rows on the Q3_A gate/up arm instead of the 128-row NVFP4 tile (1.03x concurrent decode @c4) rtx5090-required #1106: Pull request #1057 edited by widecloud
8s
8s
perf(muse): keep four packed rows on the Q3_A gate/up arm instead of the 128-row NVFP4 tile (1.03x concurrent decode @c4) rtx5090-required #1105: Pull request #1057 opened by widecloud
9s
9s
perf(muse): keep four packed rows on the Q3_A gate/up arm instead of the 128-row NVFP4 tile (1.03x concurrent decode @c4) build-attested-binaries #1122: Pull request #1057 opened by widecloud
In progress widecloud:perf/muse-gu-gemm-crossover-c4
widecloud:perf/muse-gu-gemm-crossover-c4
In progress
perf(muse): keep four packed rows on the Q3_A gate/up arm instead of the 128-row NVFP4 tile (1.03x concurrent decode @c4) sensitive-paths-guard #1747: Pull request #1057 opened by widecloud
6s
6s
perf(muse): keep four packed rows on the Q3_A gate/up arm instead of the 128-row NVFP4 tile (1.03x concurrent decode @c4) copycat-guard #761: Pull request #1057 opened by widecloud
11s
11s
perf(muse): keep four packed rows on the Q3_A gate/up arm instead of the 128-row NVFP4 tile (1.03x concurrent decode @c4) cap-open-prs #857: Pull request #1057 opened by widecloud
8s
8s
perf(moe): precompute Q4_K scales for Muse concurrent decode rtx5090-required #1104: Pull request #1055 edited by shinyuser2345-svg
10s
10s
perf(moe): precompute Q4_K scales for Muse concurrent decode build-attested-binaries #1121: Pull request #1056 opened by shinyuser2345-svg
10m 27s shinyuser2345-svg:perf/q4k-down-mma-precomputed-scales
shinyuser2345-svg:perf/q4k-down-mma-precomputed-scales
10m 27s
perf(moe): precompute Q4_K scales for Muse concurrent decode sensitive-paths-guard #1746: Pull request #1056 opened by shinyuser2345-svg
6s
6s
perf(moe): precompute Q4_K scales for Muse concurrent decode copycat-guard #760: Pull request #1056 opened by shinyuser2345-svg
10s
10s
perf(moe): precompute Q4_K scales for Muse concurrent decode cap-open-prs #856: Pull request #1056 opened by shinyuser2345-svg
9s
9s
perf(moe): precompute Q4_K scales for Muse concurrent decode rtx5090-required #1103: Pull request #1056 opened by shinyuser2345-svg
9s
9s
perf(moe): precompute Q4_K scales for Muse concurrent decode rtx5090-required #1102: Pull request #1055 edited by shinyuser2345-svg
8s
8s
perf(moe): precompute Q4_K scales for Muse concurrent decode rtx5090-required #1101: Pull request #1055 edited by shinyuser2345-svg
42s
42s
perf(moe): precompute Q4_K scales for Muse concurrent decode cap-open-prs #855: Pull request #1055 opened by shinyuser2345-svg
7s
7s
perf(moe): precompute Q4_K scales for Muse concurrent decode sensitive-paths-guard #1745: Pull request #1055 opened by shinyuser2345-svg
7s
7s
perf(moe): precompute Q4_K scales for Muse concurrent decode copycat-guard #759: Pull request #1055 opened by shinyuser2345-svg
9s
9s
perf(moe): precompute Q4_K scales for Muse concurrent decode build-attested-binaries #1120: Pull request #1055 opened by shinyuser2345-svg
10m 45s shinyuser2345-svg:perf/q4k-down-mma-precomputed-scales
shinyuser2345-svg:perf/q4k-down-mma-precomputed-scales
10m 45s
perf(moe): precompute Q4_K scales for Muse concurrent decode rtx5090-required #1100: Pull request #1055 opened by shinyuser2345-svg
10s
10s
perf(muse): let the continuous-batch ramp widen at every concurrency,… build-attested-binaries #1119: Commit bc0e47b pushed by skyrocket2026
12m 45s main
main
12m 45s
pages build and deployment pages-build-deployment #1125: by skyrocket2026
44s main
main
44s
close-stale-prs close-stale-prs #62: Scheduled
11s main
main
11s
perf(muse): let the continuous-batch ramp widen at every concurrency, not just sixteen rows and up (1.03x concurrent decode @c8) rtx5090-required #1099: Pull request #1054 edited by widecloud
8s
8s
perf(muse): let the continuous-batch ramp widen at every concurrency, not just sixteen rows and up (1.03x concurrent decode @c8) rtx5090-required #1098: Pull request #1054 edited by widecloud
7s
7s
perf(muse): let the continuous-batch ramp widen at every concurrency, not just sixteen rows and up (1.03x concurrent decode @c8) build-attested-binaries #1118: Pull request #1054 opened by widecloud
10m 56s widecloud:perf/muse-cb-ramp-gate
widecloud:perf/muse-cb-ramp-gate
10m 56s