Skip to content

Actions: PrismML-Eng/llama.cpp

Actions

Pull Request Labeler

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
labeler.yml
268 workflow runs
268 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
speculative: stop plain drafter+target falling back to KV checkpoints Pull Request Labeler #270: Pull request #173 synchronize by bri-prism
25s
25s
speculative: give the draft context the target's recurrent-state slots Pull Request Labeler #269: Pull request #174 opened by bri-prism
1m 4s
1m 4s
speculative: stop plain drafter+target falling back to KV checkpoints Pull Request Labeler #268: Pull request #173 opened by bri-prism
51s
51s
dflash: AngelSpec DFly drafter support Pull Request Labeler #267: Pull request #172 opened by bri-prism
24s
24s
cuda: enable the Hopper wgmma Q1_0/PQ2_0 prefill path by default Pull Request Labeler #266: Pull request #171 synchronize by bri-prism
1m 5s
1m 5s
cuda: enable the Hopper wgmma Q1_0/PQ2_0 prefill path by default Pull Request Labeler #265: Pull request #171 reopened by bri-prism
25s
25s
cuda: enable the Hopper wgmma Q1_0/PQ2_0 prefill path by default Pull Request Labeler #264: Pull request #171 opened by bri-prism
40s
40s
cuda: enable the Hopper wgmma Q1_0/PQ2_0 prefill path by default Pull Request Labeler #263: Pull request #169 synchronize by bri-prism
36s
36s
cuda: enable the Hopper wgmma Q1_0/PQ2_0 prefill path by default Pull Request Labeler #262: Pull request #169 opened by bri-prism
28s
28s
cuda: keep the PTX L2 prefetch hints off HIP and MUSA Pull Request Labeler #261: Pull request #168 opened by khosravipasha
27s
27s
cuda: optimize Bonsai low-bit kernels for GB10 Pull Request Labeler #260: Pull request #135 synchronize by khosravipasha
39s
39s
cuda: block-per-row FWHT for widths 512 and up (+17% / +11% decode on folded bands) Pull Request Labeler #259: Pull request #166 synchronize by bri-prism
30s
30s
cuda: block-per-row FWHT for widths 512 and up (+17% / +11% decode on folded bands) Pull Request Labeler #258: Pull request #166 opened by bri-prism
35s
35s
graph: six fusions for the hybrid-attention decode graph (-25% ops/token; Metal 2B +7.7%, 27B +2.6% decode; CUDA 9B +2.5%) Pull Request Labeler #257: Pull request #165 opened by bri-prism
24s
24s
cuda: run the PTQ1_0 MMQ path at every batch (fixes PTQ1_0 CUDA certification), env var to cap it Pull Request Labeler #256: Pull request #164 opened by bri-prism
58s
58s
cuda: query the device type with cudaDeviceGetAttribute, not cudaGetDeviceProperties (+8% decode) Pull Request Labeler #255: Pull request #163 opened by bri-prism
29s
29s
metal: fix the PTQ1_0 CPY crash and add the missing PTQ1_0 dequant-copy kernels Pull Request Labeler #254: Pull request #162 opened by bri-prism
27s
27s
metal: run every folded Hadamard width on the threadgroup FWHT kernel (+4-36% decode) Pull Request Labeler #253: Pull request #161 synchronize by bri-prism
21s
21s
metal: TQ1_0 kernel performance (dense decode +39-55%, prefill +78-90%) Pull Request Labeler #252: Pull request #156 synchronize by bri-prism
29s
29s
metal: run every folded Hadamard width on the threadgroup FWHT kernel (+4-36% decode) Pull Request Labeler #251: Pull request #161 opened by bri-prism
26s
26s
cuda: native PTQ1_0 MMQ tile loader (closes the dequantize-to-cuBLAS prefill fallback) Pull Request Labeler #250: Pull request #160 opened by bri-prism
27s
27s
metal: TQ1_0 kernel performance (dense decode +39-55%, prefill +78-90%) Pull Request Labeler #249: Pull request #156 synchronize by bri-prism
25s
25s
metal: PQ2_0 and Q2_0 mat-vec dot in exact float (+5-7% decode) Pull Request Labeler #248: Pull request #159 opened by bri-prism
22s
22s
metal: PTQ1_0 mat-mul dequantize in exact float (+7% prefill) Pull Request Labeler #247: Pull request #158 opened by bri-prism
23s
23s
metal: PTQ1_0 exact-float trit extraction (+30-38% decode) and the missing per-expert mat-vec Pull Request Labeler #246: Pull request #157 synchronize by khosravipasha
34s
34s