Skip to content

Actions: val1813/kaiwu

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
53 workflow runs
53 workflow runs

Filter by Workflow

⌥ + click/return to exclude

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: VRAM detection inflated by Resizable BAR + MoE partial OOM guard… Release Kaiwu #51: Commit e4ed413 pushed by val1813
4m 56s v0.3.2
v0.3.2
4m 56s
docs: add v0.3.1 changelog Release Kaiwu #50: Commit f8e3dde pushed by val1813
5m 52s v0.3.1
v0.3.1
5m 52s
fix: --fit on conflicts with --cpu-moe/--n-cpu-moe (root cause of MoE… Release Kaiwu #49: Commit 0940690 pushed by val1813
5m 3s v0.3.1
v0.3.1
5m 3s
fix: --fit on conflicts with --cpu-moe/--n-cpu-moe (root cause of MoE… Release Kaiwu #48: Commit 0940690 pushed by val1813
5m 5s v0.3.0
v0.3.0
5m 5s
fix: MoE models skip -sm graph on multi-GPU (expert buffer explosion) Release Kaiwu #47: Commit 2d85db8 pushed by val1813
7m 6s v0.3.0
v0.3.0
7m 6s
fix: MoE models skip -sm graph on multi-GPU (expert buffer explosion) Release Kaiwu #46: Commit 2d85db8 pushed by val1813
7m 3s v0.2.9
v0.2.9
7m 3s
docs: add v0.2.9 changelog Release Kaiwu #45: Commit 3db7c68 pushed by val1813
4m 47s v0.2.9
v0.2.9
4m 47s
fix: -sm graph runtime detection + isLikelyOOM parameter error exclusion Release Kaiwu #44: Commit 1c92429 pushed by val1813
4m 46s v0.2.8
v0.2.8
4m 46s
feat: moe_partial mode with --n-cpu-moe for fine-grained offload Release Kaiwu #43: Commit efa1cff pushed by val1813
6m 1s v0.2.8
v0.2.8
6m 1s
feat: three-mode selection + relative threshold Release Kaiwu #42: Commit f3e3084 pushed by val1813
4m 55s v0.2.8
v0.2.8
4m 55s
feat: MTP speculative decoding + n-gram lookup + KV defrag Release Kaiwu #41: Commit 3336bfc pushed by val1813
4m 30s v0.2.7
v0.2.7
4m 30s
fix: --host flag now applies to proxy (not just llama-server) Release Kaiwu #40: Commit 342951e pushed by val1813
4m 27s v0.2.7
v0.2.7
4m 27s
fix: strip CUDA_VISIBLE_DEVICES so llama-server sees all GPUs Release Kaiwu #39: Commit 3a18143 pushed by val1813
5m 38s v0.2.7
v0.2.7
5m 38s
fix: MoE models always use moe_offload (expert compute buffer OOM) Release Kaiwu #38: Commit cca04d2 pushed by val1813
4m 40s v0.2.7
v0.2.7
4m 40s
feat: Blackwell JIT warmup — permanent fix for startup timeout Release Kaiwu #37: Commit 5ac3072 pushed by val1813
5m 29s v0.2.7
v0.2.7
5m 29s
fix: skip --kv-unified on multi-GPU (KV cache lands on GPU 0 only) Release Kaiwu #36: Commit 893506d pushed by val1813
4m 58s v0.2.6
v0.2.6
4m 58s
fix: skip --kv-unified on multi-GPU (KV cache lands on GPU 0 only) Release Kaiwu #35: Commit 893506d pushed by val1813
4m 20s v0.2.5
v0.2.5
4m 20s
fix: Fingerprint panic on empty ComputeCap (P6000 and older GPUs) Release Kaiwu #34: Commit bbf1971 pushed by val1813
5m 41s v0.2.4
v0.2.4
5m 41s
docs: add v0.2.3 changelog Release Kaiwu #33: Commit e30acab pushed by val1813
5m 13s v0.2.3
v0.2.3
5m 13s
fix: Blackwell startup timeout 90s→180s, distinguish from OOM Release Kaiwu #32: Commit 1eae58e pushed by val1813
5m 21s v0.2.3
v0.2.3
5m 21s
fix: Blackwell startup timeout 90s→180s, distinguish from OOM Release Kaiwu #31: Commit 1eae58e pushed by val1813
4m 44s v0.2.2
v0.2.2
4m 44s
docs: update v0.2.2 changelog with MoE VRAM fix Release Kaiwu #30: Commit e9192c0 pushed by val1813
4m 42s v0.2.2
v0.2.2
4m 42s
fix: MoE VRAM reserve from hardcoded 1536MB to dynamic calculation Release Kaiwu #29: Commit defcad6 pushed by val1813
5m 23s v0.2.2
v0.2.2
5m 23s
docs: add v0.2.2 changelog Release Kaiwu #28: Commit 40d7583 pushed by val1813
4m 22s v0.2.2
v0.2.2
4m 22s
feat: add --host flag for LAN access Release Kaiwu #27: Commit 2ddd447 pushed by val1813
5m 7s v0.2.2
v0.2.2
5m 7s