CoolFace
Modelpublic

matbee/LFM2.5-Audio-1.5B-ONNX-cuda-gqa

sourceHugging Faceotherupdated 5mo agoView on Hugging Face
0likes6downloads
6 commits on main
eb35a145mo ago

README: full perf benchmarks (Q4 / hybrid / FP16) on RTX 4090 + Memcpy breakdown + VRAM-based config guide

matbee
e195ddf5mo ago

Patch Q4 depthformer variants + rebuild Q4 unrolled_sample (clears cos/sin, sets do_rotary=0; explains residual Q4 MatMul CPU fallback)

matbee
e42f1455mo ago

Add reproduction scripts (patch_depthformer_gqa.py + unrolled builders) and rebuild instructions

matbee
62daf875mo ago

Add unrolled+sample variants (in-graph Gumbel-max top-k), update NOTICE/README

matbee
355b9515mo ago

Add files using upload-large-folder tool

matbee
144af1d5mo ago

initial commit

matbee