Skip to content

Actions: txus/forward

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
58 workflow runs
58 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

Specialize GQA path to prepare for fused attn kernel CI #58: Commit f7a2a83 pushed by txus
2m 11s main
main
2m 11s
Specialize GQA path to prepare for fused attn kernel CI #57: Commit a040786 pushed by txus
3m 3s main
main
3m 3s
Fused Rope = 110ms to 8ms CI #56: Commit 5e83b84 pushed by txus
2m 5s main
main
2m 5s
Fused Rope = 110ms to 8ms CI #55: Commit ccbb531 pushed by txus
2m 23s main
main
2m 23s
GH actions build CI #54: Commit 4651262 pushed by txus
5m 51s main
main
5m 51s
GH actions build CI #53: Commit 758bbc8 pushed by txus
6m 10s main
main
6m 10s
GH actions build CI #52: Commit d971604 pushed by txus
4m 12s main
main
4m 12s
GH actions build CI #51: Commit 2b840bd pushed by txus
3m 53s main
main
3m 53s
GH actions build CI #50: Commit 63d1b6e pushed by txus
2m 55s main
main
2m 55s
Llama generates tokens on CUDA! CI #49: Commit 42df051 pushed by txus
1m 35s main
main
1m 35s
Llama generates tokens on CUDA! CI #48: Commit cb92d4e pushed by txus
1m 30s main
main
1m 30s
Tensor cuda is complete, and nn cuda is too CI #47: Commit bc784df pushed by txus
2m 5s main
main
2m 5s
slice is the opposite of cat CI #46: Commit 0129cb1 pushed by txus
1m 59s main
main
1m 59s
tril kernel CI #45: Commit 94cbe8d pushed by txus
1m 25s main
main
1m 25s
Arange is now like the other kernels CI #44: Commit e5b7ca3 pushed by txus
1m 27s main
main
1m 27s
Implement all elementwise operations as maps and zips CI #43: Commit 31985c7 pushed by txus
1m 35s main
main
1m 35s
Concat kernel CI #42: Commit efdd75b pushed by txus
1m 26s main
main
1m 26s
Masked fill with broadcasting CI #41: Commit d0bef3b pushed by txus
1m 30s main
main
1m 30s
Argmax CI #40: Commit bea438e pushed by txus
1m 30s main
main
1m 30s
Sum and max reductions are faaasttt CI #39: Commit 66e4b82 pushed by txus
1m 44s main
main
1m 44s
Benchmarks CI #38: Commit c93fb1f pushed by txus
1m 25s main
main
1m 25s
replace_from_ is just a cudaMemCpy CI #37: Commit 3f039c7 pushed by txus
1m 34s main
main
1m 34s
Elementwise div fp32 and mul bf16, all with vectorized loads and writes CI #36: Commit a3c6650 pushed by txus
1m 38s main
main
1m 38s
Vectorized vector subtraction in fp32 CI #35: Commit b263936 pushed by txus
2m 0s main
main
2m 0s
Add two bf16 vectors MUCH FAST CI #34: Commit e904536 pushed by txus
1m 18s main
main
1m 18s