Skip to content

Actions: hvasconcelos/libmlxforge

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
66 workflow runs
66 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
AMALIA-9B (European Portuguese) support: SPM Metaspace prepend, ChatM… CI #66: Commit 6e60b8d pushed by hvasconcelos
22m 56s master
master
22m 56s
AMALIA-9B (European Portuguese) support: SPM Metaspace prepend, ChatML detection, bf16 compute CI #65: Pull request #44 opened by hvasconcelos
27m 8s amalia-9b-support
amalia-9b-support
27m 8s
doc: prefill/decode performance technique survey (June 2026) CI #64: Pull request #43 opened by hvasconcelos
18m 14s claude/mlx-forge-performance-research-zcml6p
claude/mlx-forge-performance-research-zcml6p
18m 14s
cli: schematic — interactive HTML architecture infographic from tenso… CI #63: Commit 9960cae pushed by hvasconcelos
1h 48m 26s master
master
1h 48m 26s
cli: schematic — interactive HTML architecture infographic from tensor metadata CI #62: Pull request #42 synchronize by hvasconcelos
1h 37m 58s feature/cli-schematic
feature/cli-schematic
1h 37m 58s
cli: schematic — interactive HTML architecture infographic from tensor metadata CI #61: Pull request #42 opened by hvasconcelos
9m 27s feature/cli-schematic
feature/cli-schematic
9m 27s
Static YaRN/linear RoPE scaling + engine-level override (ABI v10) (#41) CI #60: Commit 517b40f pushed by hvasconcelos
55m 34s master
master
55m 34s
Static YaRN/linear RoPE long-context scaling + engine override (ABI v10) CI #59: Pull request #41 opened by hvasconcelos
3h 4m 40s yarn-rope-scaling
yarn-rope-scaling
3h 4m 40s
Release 0.2.0 CI #58: Commit 1b47685 pushed by hvasconcelos
1h 0m 6s master
master
1h 0m 6s
skinny_mm: simdgroup-matrix MMA kernel for big-weight decode matmuls … CI #57: Commit 5b8958e pushed by hvasconcelos
20m 5s master
master
20m 5s
skinny_mm: simdgroup-matrix MMA kernel for big-weight decode matmuls (M 5–32) CI #56: Pull request #40 synchronize by hvasconcelos
2h 4m 18s skinny-mm-mma
skinny-mm-mma
2h 4m 18s
skinny_mm: simdgroup-matrix MMA kernel for big-weight decode matmuls (M 5–32) CI #55: Pull request #40 opened by hvasconcelos
2h 1m 5s skinny-mm-mma
skinny-mm-mma
2h 1m 5s
Multi-row GEMV decode kernels — closes the batched-decode bandwidth cliff (skinny_mm, ABI v9) CI #54: Pull request #39 opened by hvasconcelos
1h 20m 41s skinny-mm-experiment
skinny-mm-experiment
1h 20m 41s
Chunked-prefill interleaving as a default-on engine feature (prefill_chunk, ABI v8) CI #53: Pull request #38 opened by hvasconcelos
1h 28m 31s chunked-prefill-default
chunked-prefill-default
1h 28m 31s
worker: chunked-prefill interleaving — +25-35% batched throughput, up to 60% lower TTFT under load CI #52: Pull request #37 synchronize by hvasconcelos
43m 23s chunked-prefill-experiment
chunked-prefill-experiment
43m 23s
worker: chunked-prefill interleaving — +25-35% batched throughput, up to 60% lower TTFT under load CI #51: Pull request #37 opened by hvasconcelos
1m 10s chunked-prefill-experiment
chunked-prefill-experiment
1m 10s
Cross-engine benchmark harness + SSE stream-termination fix CI #50: Pull request #36 synchronize by hvasconcelos
1h 53m 44s bench-harness
bench-harness
1h 53m 44s
Cross-engine benchmark harness + SSE stream-termination fix CI #49: Pull request #36 opened by hvasconcelos
10m 18s bench-harness
bench-harness
10m 18s
Prefix cache + paged (block-pool) KV storage with SSD spill tier CI #48: Pull request #35 opened by hvasconcelos
1h 47m 44s prefix-cache-paged-kv
prefix-cache-paged-kv
1h 47m 44s
Add KV-cache quantization (kv_bits 8|4), golden-gated against mlx-lm CI #47: Pull request #34 opened by hvasconcelos
1h 32m 35s kv-cache-quantization
kv-cache-quantization
1h 32m 35s
doc: add mlxforge technical white paper (LaTeX) CI #46: Pull request #33 opened by hvasconcelos
1h 42m 42s doc/whitepaper
doc/whitepaper
1h 42m 42s
Add per-token logprobs output (OpenAI logprobs/top_logprobs) CI #45: Pull request #32 synchronize by hvasconcelos
1h 36m 15s feat/logprobs-output
feat/logprobs-output
1h 36m 15s
server + ABI + bindings + batched serving: full chat history, prompt_tokens, multiple images, and decode-batched vision CI #44: Pull request #31 synchronize by hvasconcelos
1h 32m 8s work/post-qwen3-vl
work/post-qwen3-vl
1h 32m 8s
server + ABI + bindings + batched serving: full chat history, prompt_tokens, multiple images, and decode-batched vision CI #43: Pull request #31 synchronize by hvasconcelos
27m 49s work/post-qwen3-vl
work/post-qwen3-vl
27m 49s
server + ABI + bindings + batched serving: full chat history, prompt_tokens, multiple images, and decode-batched vision CI #42: Pull request #31 synchronize by hvasconcelos
5m 10s work/post-qwen3-vl
work/post-qwen3-vl
5m 10s