moonshine-base-ONNX / config.json
Xenova's picture
Xenova HF staff
Upload improved quantized ONNX weights (#2)
02858c8 verified
raw
history blame contribute delete
922 Bytes
{
"architectures": [
"MoonshineForConditionalGeneration"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 1,
"decoder_hidden_act": "silu",
"decoder_num_attention_heads": 8,
"decoder_num_hidden_layers": 8,
"decoder_num_key_value_heads": 8,
"decoder_start_token_id": 1,
"encoder_hidden_act": "gelu",
"encoder_num_attention_heads": 8,
"encoder_num_hidden_layers": 8,
"encoder_num_key_value_heads": 8,
"eos_token_id": 2,
"hidden_size": 416,
"initializer_range": 0.02,
"intermediate_size": 1664,
"is_encoder_decoder": true,
"max_position_embeddings": 512,
"model_type": "moonshine",
"num_attention_heads": 8,
"num_hidden_layers": 8,
"num_key_value_heads": 8,
"partial_rotary_factor": 0.62,
"rope_scaling": null,
"rope_theta": 10000.0,
"torch_dtype": "float32",
"transformers_version": "4.48.0.dev0",
"use_cache": true,
"vocab_size": 32768
}