{ "encoder_dim": 320, "decoder_dim": 320, "depth": 6, "nheads": 8, "head_dim": 40, "vocab_size": 32768, "bos_id": 1, "eos_id": 2, "frame_len": 80, "total_lookahead": 16, "d_model_frontend": 320, "c1": 640, "c2": 320, "frontend_state_shapes": { "sample_buffer": [ 1, 79 ], "sample_len": [ 1 ], "conv1_buffer": [ 1, 320, 4 ], "conv2_buffer": [ 1, 640, 4 ], "frame_count": [ 1 ] } }