Update README: mixed-precision (8/4) at 92.2% recommended

Add mixed 8-bit attn + 4-bit MLP variant (92.2%)

Add mixed 8-bit attn + 4-bit MLP variant (92.2%)

Add mixed 8-bit attn + 4-bit MLP variant (92.2%)

Add mixed 8-bit attn + 4-bit MLP variant (92.2%)

Add mixed 8-bit attn + 4-bit MLP variant (92.2%)

Add mixed 8-bit attn + 4-bit MLP variant (92.2%)

Add README

Add inference.py

Add recording.wav

Add obstructed.wav

Add noise.wav

Add mixed.wav

Add far_field.wav

Add echo.wav

Add dropout.wav

Add distortion.wav

Add preprocessor_config.json

Add chat_template.json

Add merges.txt

Add vocab.json

Add tokenizer_config.json

Add audio_encoder_fp32.onnx

Add generation_config.json

Add tokenizer_config.json

Add tokenizer.json

Add config.json

Add model.safetensors.index.json

Add model.safetensors

initial commit