Commit History

v2: max_total_len 512 (~20% faster), fp16 embedding (-622 MB), add int4 LM tier
c5002c0
verified

Masterx commited on

ARK-ASR-3B int8 ONNX export (0.6B-compatible layout)
b6ee80a
verified

Masterx commited on

initial commit
52de603
verified

Masterx commited on