Skip to content

Actions: quantumaikr/quant.cpp

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
671 workflow runs
671 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
feat: tq_batched_matmul_q4 — foundation for batched prefill CI #518: Commit ed4b087 pushed by unamedkr
2m 37s main
main
2m 37s
bench: prefill throughput script + 40× gap discovery CI #517: Commit 84ea97c pushed by unamedkr
2m 41s main
main
2m 41s
docs: throughput benchmark vs llama.cpp + session perf summary CI #516: Commit 012ed22 pushed by unamedkr
2m 44s main
main
2m 44s
perf(q4k): 2-row ILP in q4k_int_dot_worker CI #515: Commit aadd059 pushed by unamedkr
2m 30s main
main
2m 30s
fix(metal): Phi-3.5 Q4_K_M garbage output under default build CI #514: Commit 30dca7a pushed by unamedkr
4m 46s main
main
4m 46s
test: regression guards for today's DeltaNet fix + long-seq stress + … CI #513: Commit 361c40f pushed by unamedkr
4m 21s main
main
4m 21s
feat(cli): filter chat-template control tokens from print_token CI #512: Commit 8cea571 pushed by unamedkr
10m 30s main
main
10m 30s
feat(cli): --chat flag now correctly routes to per-family chat templates CI #511: Commit f140654 pushed by unamedkr
5m 16s main
main
5m 16s
test: relax Llama 3.1 8B check — raw '2+2=' is borderline CI #510: Commit 1e8698b pushed by unamedkr
4m 34s main
main
4m 34s
fix(server): revert Qwen3.5 think block injection (3/7 → 5/7) CI #509: Commit 53386d2 pushed by unamedkr
4m 55s main
main
4m 55s
fix(qwen35): suppress <think> token — Qwen3.5-4B short prompts now wo… CI #508: Commit ba8a615 pushed by unamedkr
4m 45s main
main
4m 45s
fix(deltanet): restore L2 norm (removal causes output collapse) + dec… CI #507: Commit 53b3323 pushed by unamedkr
4m 39s main
main
4m 39s
fix(deltanet): align decay formula with llama.cpp (#95) CI #506: Commit d26ca5e pushed by unamedkr
4m 34s main
main
4m 34s
fix: restore Phi-3.5 as RLV default (Qwen3.5-4B: 6/20 on large-doc) CI #505: Commit 7e2ca31 pushed by unamedkr
5m 30s main
main
5m 30s
feat: Qwen3.5-4B Acme 6/7 — auto-restart + model-agnostic prompts CI #504: Commit 3ad0b80 pushed by unamedkr
4m 40s main
main
4m 40s
feat: model-agnostic RLV prompts — works with Phi-3.5, Qwen3.5, Qwen3 CI #503: Commit 13dc631 pushed by unamedkr
4m 42s main
main
4m 42s
analysis: Qwen3.5-4B scored 2/7 on Acme — Phi-3.5 remains best for RLV CI #502: Commit a64e8de pushed by unamedkr
5m 48s main
main
5m 48s
fix(qwen35): early SSM probe fixes DeltaNet layer detection — Qwen3.5… CI #501: Commit e12fcbd pushed by unamedkr
4m 38s main
main
4m 38s
feat(gemma4): E4B support + comprehensive numeric analysis CI #500: Commit b8a27d2 pushed by unamedkr
5m 12s main
main
5m 12s
feat: disable Qwen3 thinking mode by default (/no_think) CI #499: Commit bbb9159 pushed by unamedkr
6m 6s main
main
6m 6s
feat: quant_generate continues from loaded KV cache (#83) CI #498: Commit e273f2b pushed by unamedkr
5m 31s main
main
5m 31s
fix(gemma4): numeric comparison with MLX-LM — divergence after layer 0 CI #497: Commit 2899fb8 pushed by unamedkr
4m 43s main
main
4m 43s
feat(default): switch to Qwen3-4B — 2.4x faster + best quality CI #496: Commit 6ea3215 pushed by unamedkr
4m 34s main
main
4m 34s
fix(gemma4): token injection + confirmed forward pass bug CI #495: Commit 91606ce pushed by unamedkr
4m 36s main
main
4m 36s
fix: load_context resets chat state + KV cache saves 57 tokens (#83) CI #494: Commit ffe668f pushed by unamedkr
4m 39s main
main
4m 39s