Commit
•
1dd2747
1
Parent(s):
388f6f5
Audio2Hero Model
Browse files- config.json +46 -0
- generation_config.json +33 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "./cache/model",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"Pop2PianoForConditionalGeneration"
|
| 5 |
+
],
|
| 6 |
+
"composer_vocab_size": 21,
|
| 7 |
+
"d_ff": 2048,
|
| 8 |
+
"d_kv": 64,
|
| 9 |
+
"d_model": 512,
|
| 10 |
+
"dataset_input_length": 1024,
|
| 11 |
+
"dataset_mel_is_conditioned": true,
|
| 12 |
+
"dataset_n_bars": 2,
|
| 13 |
+
"dataset_sampling_rate": 22050,
|
| 14 |
+
"dataset_target_length": 256,
|
| 15 |
+
"dataset_use_mel": true,
|
| 16 |
+
"decoder_start_token_id": 0,
|
| 17 |
+
"dense_act_fn": "relu",
|
| 18 |
+
"dropout_rate": 0.1,
|
| 19 |
+
"eos_token_id": 1,
|
| 20 |
+
"f_min": 10.0,
|
| 21 |
+
"feed_forward_proj": "gated-gelu",
|
| 22 |
+
"hidden_size": 512,
|
| 23 |
+
"hop_length": 1024,
|
| 24 |
+
"initializer_factor": 1.0,
|
| 25 |
+
"is_encoder_decoder": true,
|
| 26 |
+
"is_gated_act": true,
|
| 27 |
+
"layer_norm_epsilon": 1e-06,
|
| 28 |
+
"model_type": "pop2piano",
|
| 29 |
+
"n_fft": 4096,
|
| 30 |
+
"n_mels": 512,
|
| 31 |
+
"n_positions": 1024,
|
| 32 |
+
"num_attention_heads": 8,
|
| 33 |
+
"num_decoder_layers": 6,
|
| 34 |
+
"num_heads": 8,
|
| 35 |
+
"num_hidden_layers": 6,
|
| 36 |
+
"num_layers": 6,
|
| 37 |
+
"output_past": true,
|
| 38 |
+
"pad_token_id": 0,
|
| 39 |
+
"relative_attention_max_distance": 128,
|
| 40 |
+
"relative_attention_num_buckets": 32,
|
| 41 |
+
"tie_word_embeddings": false,
|
| 42 |
+
"torch_dtype": "float32",
|
| 43 |
+
"transformers_version": "4.40.1",
|
| 44 |
+
"use_cache": true,
|
| 45 |
+
"vocab_size": 2400
|
| 46 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"composer_to_feature_token": {
|
| 3 |
+
"composer1": 2052,
|
| 4 |
+
"composer10": 2061,
|
| 5 |
+
"composer11": 2062,
|
| 6 |
+
"composer12": 2063,
|
| 7 |
+
"composer13": 2064,
|
| 8 |
+
"composer14": 2065,
|
| 9 |
+
"composer15": 2066,
|
| 10 |
+
"composer16": 2067,
|
| 11 |
+
"composer17": 2068,
|
| 12 |
+
"composer18": 2069,
|
| 13 |
+
"composer19": 2070,
|
| 14 |
+
"composer2": 2053,
|
| 15 |
+
"composer20": 2071,
|
| 16 |
+
"composer21": 2072,
|
| 17 |
+
"composer3": 2054,
|
| 18 |
+
"composer4": 2055,
|
| 19 |
+
"composer5": 2056,
|
| 20 |
+
"composer6": 2057,
|
| 21 |
+
"composer7": 2058,
|
| 22 |
+
"composer8": 2059,
|
| 23 |
+
"composer9": 2060
|
| 24 |
+
},
|
| 25 |
+
"decoder_start_token_id": 0,
|
| 26 |
+
"eos_token_id": 1,
|
| 27 |
+
"max_length": 256,
|
| 28 |
+
"min_new_tokens": 50,
|
| 29 |
+
"output_logits": true,
|
| 30 |
+
"pad_token_id": 0,
|
| 31 |
+
"return_dict_in_generate": true,
|
| 32 |
+
"transformers_version": "4.40.1"
|
| 33 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bab4781f6f20aefbdd3829f425b5be802e58b9bb0ebe800856966d8147fe181d
|
| 3 |
+
size 236450544
|