Dylan M. Taylor 12fcc9fda6 Release 0.21.1

2026-11-04 19:47:19 -05:00

..

gpu

Release 0.21.1

2026-11-04 19:47:19 -05:00

decoder.rs

S2.60: parallel exact GELU: encode 3.4 s -> 2.6 s per frame

2026-10-30 08:43:31 +01:00

gpu.rs

Release 0.21.1

2026-11-04 19:47:19 -05:00

hiera.rs

S2.64: SAM 2.1 Small, Base+ and Large, each checked against PyTorch

2026-10-30 08:43:31 +01:00

lib.rs

S2.65: help and bench follow all four SAM 2 sizes; tests for the size configs and detection

2026-10-30 08:43:31 +01:00

mask.rs

Roto Brush: discard failed tracks and bound SAM2 caches (#1539)

2026-10-30 08:43:33 +01:00

memattn.rs

S2.63: GPU memory attention and memory encoder; one shared device; CPU fallback

2026-10-30 08:43:31 +01:00

memory.rs

S2.63: GPU memory attention and memory encoder; one shared device; CPU fallback

2026-10-30 08:43:31 +01:00

model.rs

S2.64: SAM 2.1 Small, Base+ and Large, each checked against PyTorch

2026-10-30 08:43:31 +01:00

neck.rs

Roto Brush: validate GPU layouts and align object outlines (#1539)

2026-10-30 08:43:33 +01:00

nn.rs

S2.61: parallel bias add fused with GELU; GPU matmul spike

2026-10-30 08:43:31 +01:00

prompt.rs

#1539: match reference box padding and use a flat GPU module

2026-10-30 08:43:33 +01:00

resize.rs

Roto Brush: discard failed tracks and bound SAM2 caches (#1539)

2026-10-30 08:43:33 +01:00

session.rs

#1539: validate benchmark inputs and preserve shared refinement strokes

2026-10-30 10:00:33 +01:00

track.rs

Roto Brush: discard failed tracks and bound SAM2 caches (#1539)

2026-10-30 08:43:33 +01:00

weights.rs

S2.64: SAM 2.1 Small, Base+ and Large, each checked against PyTorch

2026-10-30 08:43:31 +01:00