Skip to content

Actions: mudler/vllm.cpp

Actions

ci

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
ci.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
perf(ROCm): replace software Dp4a with hardware v_dot4_i32_iu8 ci #6635: Pull request #2363 synchronize by ghazni101
Pending ghazni101:row/ROCM-HW-DP4A
ghazni101:row/ROCM-HW-DP4A
Pending
merge(MODEL-DFLASH2-EXL3): load and run a DFlash2 draft that ships tr… ci #6634: Commit dd2123f pushed by mudler
Pending main
main
Pending
ci(ENG-CI-ORACLE-GOLDENS): run the oracle-goldens digests on a lane, not only on a developer box ci #6633: Pull request #2555 closed by mudler
Queued row/ENG-CI-ORACLE-GOLDENS
row/ENG-CI-ORACLE-GOLDENS
Queued
ci(ENG-CI-ORACLE-GOLDENS): run the oracle-goldens digests on a lane, … ci #6632: Commit 6974557 pushed by mudler
24m 47s main
main
24m 47s
feat(BACKEND-ROCM): register kFusedNormRope, so GLM-5.3 stops taking the path a keep-quant MLA checkpoint cannot survive ci #6631: Pull request #2572 closed by mudler
Pending row/BACKEND-ROCM-FUSED-NORM-ROPE
row/BACKEND-ROCM-FUSED-NORM-ROPE
Pending
merge(BACKEND-ROCM-FUSED-NORM-ROPE): register kFusedNormRope on ROCm,… ci #6630: Commit 9f3e6e2 pushed by mudler
9m 1s main
main
9m 1s
feat(BACKEND-ROCM): register kFusedNormRope, so GLM-5.3 stops taking the path a keep-quant MLA checkpoint cannot survive ci #6629: Pull request #2572 synchronize by mudler
11m 4s row/BACKEND-ROCM-FUSED-NORM-ROPE
row/BACKEND-ROCM-FUSED-NORM-ROPE
11m 4s
perf(ROCm): replace software Dp4a with hardware v_dot4_i32_iu8 ci #6628: Pull request #2363 synchronize by ghazni101
1h 7m 24s ghazni101:row/ROCM-HW-DP4A
ghazni101:row/ROCM-HW-DP4A
1h 7m 24s
feat(rocm): fp8 KV cache prefill fast path via SharedK kernel ci #6627: Pull request #2361 synchronize by ghazni101
Queued ghazni101:row/fp8-prefill-fastpath
ghazni101:row/fp8-prefill-fastpath
Queued
ci ci #6626: Scheduled
In progress main
main
In progress
feat(rocm): fp8 KV cache prefill fast path via SharedK kernel ci #6625: Pull request #2361 synchronize by ghazni101
9m 45s ghazni101:row/fp8-prefill-fastpath
ghazni101:row/fp8-prefill-fastpath
9m 45s
feat(rocm): fp8 KV cache prefill fast path via SharedK kernel ci #6624: Pull request #2361 synchronize by ghazni101
19m 55s ghazni101:row/fp8-prefill-fastpath
ghazni101:row/fp8-prefill-fastpath
19m 55s
feat(rocm): fp8 KV cache prefill fast path via SharedK kernel ci #6623: Pull request #2361 synchronize by ghazni101
5m 12s ghazni101:row/fp8-prefill-fastpath
ghazni101:row/fp8-prefill-fastpath
5m 12s
diag(MODEL-MM-QWEN4-EXP): the MoE expert selections DIFFER, and a third of the boundaries are exact bf16 ties (#2552) ci #6622: Pull request #2556 synchronize by mudler
In progress row/MOEDIV-q4exp-router-2552
row/MOEDIV-q4exp-router-2552
In progress
feat(rocm): fp8 KV cache prefill fast path via SharedK kernel ci #6621: Pull request #2361 synchronize by ghazni101
23m 45s ghazni101:row/fp8-prefill-fastpath
ghazni101:row/fp8-prefill-fastpath
23m 45s
merge: origin/main into the CPU-floor flake record ci #6620: Commit aedad72 pushed by mudler
Queued main
main
Queued
perf(ROCm): replace software Dp4a with hardware v_dot4_i32_iu8 ci #6619: Pull request #2363 synchronize by ghazni101
Queued ghazni101:row/ROCM-HW-DP4A
ghazni101:row/ROCM-HW-DP4A
Queued
MODEL-MM-GLM53-FLASH W9c-3b: the --device cuda SIGSEGV is a KV-page residency defect, not a mixed-residency one ci #6618: Pull request #2483 closed by mudler
3h 5m 39s row/MODEL-MM-GLM53-FLASH-DEVSEGV
row/MODEL-MM-GLM53-FLASH-DEVSEGV
3h 5m 39s
merge(MODEL-MM-GLM53-FLASH-DEVSEGV): route KV spans through the backe… ci #6617: Commit b898d6f pushed by mudler
8m 47s main
main
8m 47s
feat(BACKEND-ROCM): register kFusedNormRope, so GLM-5.3 stops taking the path a keep-quant MLA checkpoint cannot survive ci #6616: Pull request #2572 synchronize by mudler
Queued row/BACKEND-ROCM-FUSED-NORM-ROPE
row/BACKEND-ROCM-FUSED-NORM-ROPE
Queued
MODEL-MM-GLM53-FLASH W9c-3b: the --device cuda SIGSEGV is a KV-page residency defect, not a mixed-residency one ci #6615: Pull request #2483 synchronize by mudler
1h 44m 17s row/MODEL-MM-GLM53-FLASH-DEVSEGV
row/MODEL-MM-GLM53-FLASH-DEVSEGV
1h 44m 17s
feat(BACKEND-ROCM): register kFusedNormRope, so GLM-5.3 stops taking the path a keep-quant MLA checkpoint cannot survive ci #6614: Pull request #2572 opened by localai-org-maint-bot
1h 47m 55s row/BACKEND-ROCM-FUSED-NORM-ROPE
row/BACKEND-ROCM-FUSED-NORM-ROPE
1h 47m 55s
merge: row/BACKEND-ROCM-GFX1151-CAUSE (#2511) -- Gate 2 clean at gate… ci #6613: Commit 43bcb86 pushed by mudler
1h 56m 22s main
main
1h 56m 22s
docs(MODEL-QWEN35-GDN-EXL3): read preflight by its output, not by its… ci #6612: Commit 4fe3852 pushed by mudler
9m 25s main
main
9m 25s
perf(ROCm): replace software Dp4a with hardware v_dot4_i32_iu8 ci #6611: Pull request #2363 synchronize by ghazni101
29m 45s ghazni101:row/ROCM-HW-DP4A
ghazni101:row/ROCM-HW-DP4A
29m 45s