Skip to content

Actions: thecodacus/llama.cpp

Actions

CI (self-hosted)

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
build-self-hosted.yml
17 workflow runs
17 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
fix: report fit math when expert cache pack allocation fails CI (self-hosted) #17: Pull request #7 opened by thecodacus
1d 0h 0m 1s fable5/moe-cache-diagnostics
fable5/moe-cache-diagnostics
1d 0h 0m 1s
feat: wire MoE expert cache into laguna graph CI (self-hosted) #16: Pull request #6 opened by thecodacus
1d 0h 0m 2s fable5/moe-cache-laguna
fable5/moe-cache-laguna
1d 0h 0m 2s
feat: MoE expert cache — keep hot routed experts resident in VRAM CI (self-hosted) #15: Pull request #5 opened by thecodacus
1d 0h 0m 2s fable5/moe-expert-cache
fable5/moe-expert-cache
1d 0h 0m 2s
cohere2 moe template parser: enforce JSON schema for text responses i… CI (self-hosted) #14: Commit 8f5ab83 pushed by thecodacus
1d 0h 0m 2s master
master
1d 0h 0m 2s
feat: fused turbo MMA flash-attention decode (turbo4/3/2, GQA tensor-core path) CI (self-hosted) #13: Pull request #4 opened by thecodacus
1d 0h 0m 2s fable5/turbo-mma-decode
fable5/turbo-mma-decode
1d 0h 0m 2s
fix: allow mixed turbo/q8_0 KV cache types in CUDA flash-attention selection CI (self-hosted) #12: Pull request #3 synchronize by thecodacus
1d 0h 0m 2s fable5/turbo-fattn-vec-routing
fable5/turbo-fattn-vec-routing
1d 0h 0m 2s
fix: allow mixed turbo/q8_0 KV cache types in CUDA flash-attention selection CI (self-hosted) #11: Pull request #3 synchronize by thecodacus
1m 53s fable5/turbo-fattn-vec-routing
fable5/turbo-fattn-vec-routing
1m 53s
fix: allow mixed turbo/q8_0 KV cache types in CUDA flash-attention selection CI (self-hosted) #10: Pull request #3 opened by thecodacus
14m 32s fable5/turbo-fattn-vec-routing
fable5/turbo-fattn-vec-routing
14m 32s
feat: CUDA Q2_0 support (Bonsai-27B ternary/binary on Qwen3.6 arch) CI (self-hosted) #9: Pull request #2 opened by thecodacus
1d 0h 0m 2s fable5/cuda-q2_0
fable5/cuda-q2_0
1d 0h 0m 2s
feat: TurboQuant KV-cache quantization — all backends CI (self-hosted) #8: Pull request #1 synchronize by thecodacus
1d 0h 0m 3s fable5/turboquant
fable5/turboquant
1d 0h 0m 3s
feat: TurboQuant KV-cache quantization — all backends CI (self-hosted) #7: Pull request #1 synchronize by thecodacus
44m 58s fable5/turboquant
fable5/turboquant
44m 58s
feat: TurboQuant KV-cache quantization — all backends CI (self-hosted) #6: Pull request #1 synchronize by thecodacus
1h 7m 15s fable5/turboquant
fable5/turboquant
1h 7m 15s
feat: TurboQuant KV-cache quantization — all backends CI (self-hosted) #5: Pull request #1 synchronize by thecodacus
7h 52m 27s fable5/turboquant
fable5/turboquant
7h 52m 27s
feat: TurboQuant KV-cache quantization — all backends CI (self-hosted) #4: Pull request #1 synchronize by thecodacus
10m 52s fable5/turboquant
fable5/turboquant
10m 52s
feat: TurboQuant KV-cache quantization — all backends CI (self-hosted) #3: Pull request #1 opened by thecodacus
5m 33s fable5/turboquant
fable5/turboquant
5m 33s
sync : ggml CI (self-hosted) #2: Commit 7d56da7 pushed by thecodacus
18h 29m 28s master
master
18h 29m 28s
metal: add col2im_1d op (f32/f16/bf16) (#25176) CI (self-hosted) #1: Commit f36e5c3 pushed by thecodacus
1d 0h 0m 2s master
master
1d 0h 0m 2s