Skip to content

Actions: artalis-io/bitnet.c

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
602 workflow runs
602 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
Streaming callback API, WASM chat mode, and C safety audit fixes CI #27: Commit d5e9854 pushed by farkasmark
24s main
main
24s
Chat auto-reset and /reset command for multi-turn quality CI #26: Commit ec11e07 pushed by farkasmark
18s main
main
18s
Document correct chat template source (HuggingFace tokenizer_config.j… CI #25: Commit fef2b5a pushed by farkasmark
19s main
main
19s
Sliding window KV cache, PGO build target, roadmap update CI #24: Commit d964bd8 pushed by farkasmark
17s main
main
17s
Optional F16 KV cache (--kv16): halve attention DRAM bandwidth CI #23: Commit c4a49f2 pushed by farkasmark
24s main
main
24s
Batch prompt prefill: skip logits for intermediate prompt tokens CI #22: Commit d6db0f8 pushed by farkasmark
17s main
main
17s
Batch prompt prefill: skip logits for intermediate prompt tokens CI #21: Commit 401bc7a pushed by farkasmark
20s main
main
20s
Add --chat REPL mode with multi-turn KV cache reuse CI #20: Commit be10b95 pushed by farkasmark
21s main
main
21s
Flash attention (--flash), README update, design decisions CI #19: Commit be64b64 pushed by farkasmark
30s main
main
30s
INT8 output embeddings for SDOT logits kernel (~52 tok/s) CI #18: Commit 11723dc pushed by farkasmark
55s main
main
55s
INT8 output embeddings for SDOT logits kernel (~52 tok/s) CI #17: Commit 35c4040 pushed by farkasmark
16s main
main
16s
Enable native FP16 logits, add SDOT prefetch (~46 tok/s) CI #16: Commit 4d949ef pushed by farkasmark
20s main
main
20s
Integrate sh_arena + sh_log, eliminate magic constants CI #15: Commit d95d5ea pushed by farkasmark
18s main
main
18s
Add bn_ namespace prefix to all public symbols CI #14: Commit be9b13c pushed by farkasmark
19s main
main
19s
Replace OpenMP with pthread thread pool + audit hardening (~41 tok/s) CI #13: Commit 1ab9dad pushed by farkasmark
24s main
main
24s
Arithmetic ternary decode + RoPE cos/sin precompute + direct SDOT (~3… CI #12: Commit 52b5fc5 pushed by farkasmark
21s main
main
21s
Precompute RoPE frequencies + unroll GQA attention loops CI #11: Commit f8d0f9f pushed by farkasmark
18s main
main
18s
F16 native FMA for logits dot product (~34 tok/s) CI #10: Commit 864a85c pushed by farkasmark
20s main
main
20s
SDOT int8 accumulation + batch matvec (2.1x speedup to ~32.7 tok/s) CI #9: Commit 4992796 pushed by farkasmark
16s main
main
16s
8-accumulator ILP, larger prefetch, and P-core affinity CI #8: Commit f312b9a pushed by farkasmark
24s main
main
24s
Add prefetch hints for weight data in ternary_matvec() CI #7: Commit a1f31d7 pushed by farkasmark
21s main
main
21s
Add ARM NEON SIMD for transformer forward pass (5.0x total speedup) CI #6: Commit f2072d8 pushed by farkasmark
21s main
main
21s
Add ARM NEON SIMD for ternary_matvec() inner loops (3.6x speedup) CI #5: Commit 211504e pushed by farkasmark
20s main
main
20s
Enable _GNU_SOURCE on Linux for POSIX/GNU extensions CI #4: Commit 95252af pushed by farkasmark
15s main
main
15s
Enable _GNU_SOURCE on Linux for POSIX/GNU extensions CI #3: Commit a196ec7 pushed by farkasmark
18s main
main
18s
ProTip! You can narrow down the results and go further in time using created:<2026-03-12 or the other filters available.