Skip to content

Actions: lightseekorg/tokenspeed

Actions

PR Test AMD

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
pr-test-amd.yml
2,500+ workflow runs
2,500+ 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
refactor(runtime): unify eager/graph decode and spec/non-spec sampling paths PR Test AMD #3573: Pull request #1299 synchronize by FC-Li
Queued unified-decode-path
unified-decode-path
Queued
perf(qwen4): optimize gated residual kernels and PDL (#1323) PR Test AMD #3572: Commit a895243 pushed by lightseek-bot
In progress main
main
In progress
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3571: Pull request #1323 synchronize by tuanzhangCS
14m 18s zt/qwen4_opt
zt/qwen4_opt
14m 18s
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3570: Pull request #1323 synchronize by tuanzhangCS
30m 33s zt/qwen4_opt
zt/qwen4_opt
30m 33s
refactor(runtime): unify eager/graph decode and spec/non-spec sampling paths PR Test AMD #3569: Pull request #1299 synchronize by FC-Li
1h 17m 23s unified-decode-path
unified-decode-path
1h 17m 23s
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3568: Pull request #1323 synchronize by tuanzhangCS
15m 15s zt/qwen4_opt
zt/qwen4_opt
15m 15s
perf(k3): 256-bit skinny GEMV configs for the measured decode shapes PR Test AMD #3567: Pull request #1341 opened by nperrin-fr
7m 25s perf/skinny-vw16-configs
perf/skinny-vw16-configs
7m 25s
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3566: Pull request #1323 synchronize by tuanzhangCS
16m 55s zt/qwen4_opt
zt/qwen4_opt
16m 55s
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3565: Pull request #1323 ready_for_review by tuanzhangCS
17s zt/qwen4_opt
zt/qwen4_opt
17s
docs: fix kimi_k2.5 agentic bench README cd paths (#1340) PR Test AMD #3564: Commit ba1b1c7 pushed by torchspec-bot
1h 19m 48s main
main
1h 19m 48s
docs: fix kimi_k2.5 agentic bench README cd paths PR Test AMD #3563: Pull request #1340 opened by fuergaosi233
17m 31s fuergaosi233:docs/fix-kimi-k25-readme-cd-path
fuergaosi233:docs/fix-kimi-k25-readme-cd-path
17m 31s
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3562: Pull request #1323 synchronize by tuanzhangCS
24s zt/qwen4_opt
zt/qwen4_opt
24s
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3561: Pull request #1323 synchronize by tuanzhangCS
15m 26s zt/qwen4_opt
zt/qwen4_opt
15m 26s
perf(qwen4): optimize gated residual kernels and PDL PR Test AMD #3560: Pull request #1323 ready_for_review by tuanzhangCS
18s zt/qwen4_opt
zt/qwen4_opt
18s
fix(moe): pad NVFP4 expert shards for TRT-LLM (#1322) PR Test AMD #3559: Commit 0fe47ab pushed by lightseek-bot
42m 30s main
main
42m 30s
perf(k3): time the MoE tactic sweep by graph replay and floor it at a… PR Test AMD #3558: Commit 1271aad pushed by lightseek-bot
57m 14s main
main
57m 14s
perf(k3): time the MoE tactic sweep by graph replay and floor it at auto-tune PR Test AMD #3557: Pull request #1338 opened by nperrin-fr
10m 48s perf/k3-moe-tactic-graph-floor
perf/k3-moe-tactic-graph-floor
10m 48s
perf(amd): derive gfx1250 prefill occupancy cutoff from the device PR Test AMD #3556: Pull request #1337 synchronize by nithinsubbiah
1h 34m 6s nithinsubbiah:perf/gfx1250-prefill-tile-followup
nithinsubbiah:perf/gfx1250-prefill-tile-followup
1h 34m 6s
perf(amd): derive gfx1250 prefill occupancy cutoff from the device PR Test AMD #3555: Pull request #1337 synchronize by nithinsubbiah
37m 7s nithinsubbiah:perf/gfx1250-prefill-tile-followup
nithinsubbiah:perf/gfx1250-prefill-tile-followup
37m 7s
perf(amd): derive gfx1250 prefill occupancy cutoff from the device PR Test AMD #3554: Pull request #1337 synchronize by nithinsubbiah
Action required nithinsubbiah:perf/gfx1250-prefill-tile-followup
nithinsubbiah:perf/gfx1250-prefill-tile-followup
Action required
perf(amd): derive gfx1250 prefill occupancy cutoff from the device PR Test AMD #3553: Pull request #1337 opened by nithinsubbiah
Action required nithinsubbiah:perf/gfx1250-prefill-tile-followup
nithinsubbiah:perf/gfx1250-prefill-tile-followup
Action required
perf(kimi3): avoid shape-specialized AttnRes compilation (#1336) PR Test AMD #3552: Commit c3ea3cd pushed by lightseek-bot
1h 28m 18s main
main
1h 28m 18s
perf(kimi3): avoid shape-specialized AttnRes compilation PR Test AMD #3551: Pull request #1336 opened by qedawkins
8m 48s qedawkins:users/qedawkins/fix-kimi3-attnres-shape-jit
qedawkins:users/qedawkins/fix-kimi3-attnres-shape-jit
8m 48s
perf(kimi3): let the MoE tail join its reductions without a lane (#1330) PR Test AMD #3550: Commit 405acad pushed by lightseek-bot
47m 27s main
main
47m 27s
fix(engine): stop profiling without a device-side rendezvous (#1331) PR Test AMD #3549: Commit c50ea85 pushed by lightseek-bot
5m 38s main
main
5m 38s