MERaLiON-3-10B-RotorQuant-MLX-8bit / adaptor_config.json
majentik's picture
Add MLX quantized model with KV cache compression
ba8ef4c verified
raw
history blame contribute delete
170 Bytes
{
"speech_hidden_size": 1280,
"text_hidden_size": 3584,
"scale_factor": 5,
"use_projection": false,
"use_weighted_layer_sum": true,
"num_encoder_layers": 32
}