Skip to content

Actions: sauravpanda/forge-llm

Actions

Publish Python Wheels

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
publish-python.yml
44 workflow runs
44 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
feat: v0.9.13 — per-projection dtype mixing for native Q4_K_M GGUFs Publish Python Wheels #44: Commit 619607c pushed by sauravpanda
1d 0h 0m 2s v0.9.13
v0.9.13
1d 0h 0m 2s
feat: v0.9.12 — Q6_K codegen + AOT validation (closes the Q6_K arc) Publish Python Wheels #43: Commit 7f52de6 pushed by sauravpanda
1d 0h 0m 2s v0.9.12
v0.9.12
1d 0h 0m 2s
feat: v0.9.11 — native Q6_K quantizer + dot kernel (foundation) Publish Python Wheels #42: Commit 6d3ecaa pushed by sauravpanda
1d 0h 0m 1s v0.9.11
v0.9.11
1d 0h 0m 1s
feat: v0.9.10 — --force-q4k validates AOT Q4_K codegen end-to-end Publish Python Wheels #41: Commit 29a5b75 pushed by sauravpanda
1d 0h 0m 1s v0.9.10
v0.9.10
1d 0h 0m 1s
feat: v0.9.9 — AOT --score-corpus mode (codegen mathematically valida… Publish Python Wheels #40: Commit 05a54fd pushed by sauravpanda
1d 0h 0m 2s v0.9.9
v0.9.9
1d 0h 0m 2s
fix: v0.9.8 — AOT binary CLI parser eats flag values into prompt Publish Python Wheels #39: Commit 7828557 pushed by sauravpanda
1d 0h 0m 2s v0.9.8
v0.9.8
1d 0h 0m 2s
fix: v0.9.7 — f32_to_f16 round-to-nearest-even (fixes simulate-Q4_K bug) Publish Python Wheels #38: Commit 3ee634b pushed by sauravpanda
1d 0h 0m 1s v0.9.7
v0.9.7
1d 0h 0m 1s
feat: v0.9.6 — forge bench-perplexity infrastructure Publish Python Wheels #37: Commit 1ae0d32 pushed by sauravpanda
1d 0h 0m 1s v0.9.6
v0.9.6
1d 0h 0m 1s
feat: v0.9.5 — make_qkx2_quants port for Q4_K Publish Python Wheels #36: Commit d0c51a5 pushed by sauravpanda
1d 0h 0m 2s v0.9.5
v0.9.5
1d 0h 0m 2s
perf: v0.9.4 — Q4_K dot4 ILP (now faster than Q8_0) Publish Python Wheels #35: Commit d5af7fd pushed by sauravpanda
1d 0h 0m 2s v0.9.4
v0.9.4
1d 0h 0m 2s
perf: v0.9.3 — NEON Q4_K kernel (1.58× decode speedup) Publish Python Wheels #34: Commit 84e0269 pushed by sauravpanda
1d 0h 0m 2s v0.9.3
v0.9.3
1d 0h 0m 2s
feat: v0.9.2 — native Q4_K storage + scalar kernel Publish Python Wheels #33: Commit 1cd3872 pushed by sauravpanda
1d 0h 0m 2s v0.9.2
v0.9.2
1d 0h 0m 2s
feat: v0.9.1 — per-tensor lm_head dtype dispatch (infra) Publish Python Wheels #32: Commit 59a4760 pushed by sauravpanda
1d 0h 0m 1s v0.9.1
v0.9.1
1d 0h 0m 1s
feat: v0.9.0 — Q4_K_M / K-quant GGUF compatibility Publish Python Wheels #31: Commit d8bf551 pushed by sauravpanda
1d 0h 0m 1s v0.9.0
v0.9.0
1d 0h 0m 1s
docs: v0.8.8 — v0.8.5 SWA kernel validated on Phi-3 Publish Python Wheels #30: Commit c6bee4a pushed by sauravpanda
1d 0h 0m 2s v0.8.8
v0.8.8
1d 0h 0m 2s
fix: v0.8.7 — forge bench clarifies interpreter vs AOT Publish Python Wheels #29: Commit 9c0e666 pushed by sauravpanda
1d 0h 0m 2s v0.8.7
v0.8.7
1d 0h 0m 2s
feat: v0.8.6 — FORGE_BATCHED_PREFILL runtime toggle + blog addendum Publish Python Wheels #28: Commit 3580b0d pushed by sauravpanda
1d 0h 0m 2s v0.8.6
v0.8.6
1d 0h 0m 2s
perf: v0.8.5 — batched sliding-window attention Publish Python Wheels #27: Commit f2e2b7e pushed by sauravpanda
1d 0h 0m 2s v0.8.5
v0.8.5
1d 0h 0m 2s
fix: v0.8.4 — SWA correctness in CPU batched prefill Publish Python Wheels #26: Commit 223e33a pushed by sauravpanda
1d 0h 0m 2s v0.8.4
v0.8.4
1d 0h 0m 2s
docs: v0.8.3 — CPU prefill benchmarks + v0.8.x blog post Publish Python Wheels #25: Commit dcc78bf pushed by sauravpanda
1d 0h 0m 1s v0.8.3
v0.8.3
1d 0h 0m 1s
perf: v0.8.2 — Q4_0 batched prefill Publish Python Wheels #24: Commit c343865 pushed by sauravpanda
1d 0h 0m 3s v0.8.2
v0.8.2
1d 0h 0m 3s
perf: v0.8.1 — CPU batched attention (Q-tiled flash) Publish Python Wheels #23: Commit f73a3d5 pushed by sauravpanda
1d 0h 0m 2s v0.8.1
v0.8.1
1d 0h 0m 2s
perf: v0.8.0 — CPU batched prefill for Q8_0 (closes #211) Publish Python Wheels #22: Commit 826fa9f pushed by sauravpanda
1d 0h 0m 2s v0.8.0
v0.8.0
1d 0h 0m 2s
perf: v0.7.13 — MPS attention for MHA models + GQA short-prompt fix Publish Python Wheels #21: Commit ab305ea pushed by sauravpanda
1d 0h 0m 1s v0.7.13
v0.7.13
1d 0h 0m 1s
chore: release v0.7.12 — CI cleanup for Rust 1.95 clippy lints Publish Python Wheels #20: Commit 5fdd9eb pushed by sauravpanda
1d 0h 0m 2s v0.7.12
v0.7.12
1d 0h 0m 2s