Upload 5 files
Browse files- config.yaml +65 -0
- pytorch_model.bin +3 -0
- tokenizer.json +465 -0
- trainer_state.json +0 -0
- training_args.bin +3 -0
config.yaml
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data:
|
| 2 |
+
data_dir: "data/MOSES/shapemol/MOSES2_training_val_dataset.pkl" # Path to the raw data (pickle file)
|
| 3 |
+
save_dir: "./data/MOSES" # Directory to save the LMDB database
|
| 4 |
+
num_samples: 1024 # Number of point cloud samples generated per molecule
|
| 5 |
+
num_workers: 20 # Number of parallel workers for data processing
|
| 6 |
+
aug_mode: 1
|
| 7 |
+
aug_times: 2
|
| 8 |
+
map_size: 500 # LMDB map_size in bytes (100GB)
|
| 9 |
+
batch_size: 20000 # Batch size for processing data
|
| 10 |
+
seed: 42 # Random seed for reproducibility
|
| 11 |
+
use_smiles: False # Whether to use SMILES strings for data processing
|
| 12 |
+
|
| 13 |
+
model:
|
| 14 |
+
surf:
|
| 15 |
+
n: 2
|
| 16 |
+
normal_channel: True
|
| 17 |
+
mlp:
|
| 18 |
+
hidden_dim: 256 # Hidden dimension for MLP
|
| 19 |
+
output_dim: 768 # Output dimension for MLP
|
| 20 |
+
num_layers: 2 # Number of layers in MLP
|
| 21 |
+
dropout_rate: 0.1 # Dropout rate in MLP
|
| 22 |
+
activation_function: "relu" # Activation function for MLP
|
| 23 |
+
bart:
|
| 24 |
+
max_position_embeddings: 512
|
| 25 |
+
d_model: 768
|
| 26 |
+
encoder_layers: 0
|
| 27 |
+
decoder_layers: 6
|
| 28 |
+
encoder_attention_heads: 0
|
| 29 |
+
decoder_attention_heads: 8
|
| 30 |
+
encoder_ffn_dim: 0
|
| 31 |
+
decoder_ffn_dim: 3072
|
| 32 |
+
activation_function: 'gelu'
|
| 33 |
+
generation_config:
|
| 34 |
+
do_sample: true
|
| 35 |
+
max_length: 512
|
| 36 |
+
top_k: 50
|
| 37 |
+
top_p: 1.0
|
| 38 |
+
temperature: 1.0
|
| 39 |
+
num_return_sequences: 50
|
| 40 |
+
|
| 41 |
+
train:
|
| 42 |
+
output_dir: "./checkpoints/conditional/surfbartv2-sample1024-merge-angles-0421" # TODO: Directory to save model and checkpoints
|
| 43 |
+
resume_path: null
|
| 44 |
+
overwrite_output_dir: true # Whether to overwrite existing outputs
|
| 45 |
+
num_train_epochs: 50 # Total number of training epochs
|
| 46 |
+
per_device_train_batch_size: 150 # Training batch size per device
|
| 47 |
+
per_device_eval_batch_size: 4 # Evaluation batch size per device
|
| 48 |
+
dataloader_num_workers: 2 # Number of workers for data loading
|
| 49 |
+
save_total_limit: 6 # Maximum number of checkpoints to keep
|
| 50 |
+
logging_steps: 50 # Steps interval for logging
|
| 51 |
+
eval_strategy: "steps" # Evaluation strategy (e.g., "steps", "epoch")
|
| 52 |
+
eval_steps: 5000 # Evaluation frequency (in steps)
|
| 53 |
+
do_eval: true # Whether to perform evaluation
|
| 54 |
+
learning_rate: 1e-4 # Initial learning rate
|
| 55 |
+
warmup_ratio: 0.1 # Warm-up ratio for learning rate scheduler
|
| 56 |
+
save_strategy: "steps" # Save strategy (e.g., "steps", "epoch")
|
| 57 |
+
save_steps: 5000 # Save frequency (in steps)
|
| 58 |
+
load_best_model_at_end: true # Load the best model at the end of training
|
| 59 |
+
logging_first_step: true # Log the first training step
|
| 60 |
+
bf16: True # Whether to use bf16 precision
|
| 61 |
+
early_stopping_patience: 10 # Patience for early stopping callback
|
| 62 |
+
early_stopping_threshold: 0 # Threshold for early stopping
|
| 63 |
+
seed: 42
|
| 64 |
+
|
| 65 |
+
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8d76efa64cee8c6f072535db4e8f9d44844b336286c3baad75e6f5d7fe21b171
|
| 3 |
+
size 238174446
|
tokenizer.json
ADDED
|
@@ -0,0 +1,465 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"<BOS>": 0,
|
| 3 |
+
"<EOS>": 1,
|
| 4 |
+
"<PAD>": 2,
|
| 5 |
+
"<UNK>": 3,
|
| 6 |
+
"<MASK>": 4,
|
| 7 |
+
"!": 5,
|
| 8 |
+
"\"": 6,
|
| 9 |
+
"#": 7,
|
| 10 |
+
"$": 8,
|
| 11 |
+
"%": 9,
|
| 12 |
+
"&": 10,
|
| 13 |
+
"'": 11,
|
| 14 |
+
"(": 12,
|
| 15 |
+
")": 13,
|
| 16 |
+
"*": 14,
|
| 17 |
+
"+": 15,
|
| 18 |
+
",": 16,
|
| 19 |
+
"-": 17,
|
| 20 |
+
".": 18,
|
| 21 |
+
"/": 19,
|
| 22 |
+
"0": 20,
|
| 23 |
+
"1": 21,
|
| 24 |
+
"2": 22,
|
| 25 |
+
"3": 23,
|
| 26 |
+
"4": 24,
|
| 27 |
+
"5": 25,
|
| 28 |
+
"6": 26,
|
| 29 |
+
"7": 27,
|
| 30 |
+
"8": 28,
|
| 31 |
+
"9": 29,
|
| 32 |
+
":": 30,
|
| 33 |
+
";": 31,
|
| 34 |
+
"<": 32,
|
| 35 |
+
"=": 33,
|
| 36 |
+
">": 34,
|
| 37 |
+
"?": 35,
|
| 38 |
+
"@": 36,
|
| 39 |
+
"A": 37,
|
| 40 |
+
"B": 38,
|
| 41 |
+
"C": 39,
|
| 42 |
+
"D": 40,
|
| 43 |
+
"E": 41,
|
| 44 |
+
"F": 42,
|
| 45 |
+
"G": 43,
|
| 46 |
+
"H": 44,
|
| 47 |
+
"I": 45,
|
| 48 |
+
"J": 46,
|
| 49 |
+
"K": 47,
|
| 50 |
+
"L": 48,
|
| 51 |
+
"M": 49,
|
| 52 |
+
"N": 50,
|
| 53 |
+
"O": 51,
|
| 54 |
+
"P": 52,
|
| 55 |
+
"Q": 53,
|
| 56 |
+
"R": 54,
|
| 57 |
+
"S": 55,
|
| 58 |
+
"T": 56,
|
| 59 |
+
"U": 57,
|
| 60 |
+
"V": 58,
|
| 61 |
+
"W": 59,
|
| 62 |
+
"X": 60,
|
| 63 |
+
"Y": 61,
|
| 64 |
+
"Z": 62,
|
| 65 |
+
"[": 63,
|
| 66 |
+
"\\": 64,
|
| 67 |
+
"]": 65,
|
| 68 |
+
"^": 66,
|
| 69 |
+
"_": 67,
|
| 70 |
+
"`": 68,
|
| 71 |
+
"a": 69,
|
| 72 |
+
"b": 70,
|
| 73 |
+
"c": 71,
|
| 74 |
+
"d": 72,
|
| 75 |
+
"e": 73,
|
| 76 |
+
"f": 74,
|
| 77 |
+
"g": 75,
|
| 78 |
+
"h": 76,
|
| 79 |
+
"i": 77,
|
| 80 |
+
"j": 78,
|
| 81 |
+
"k": 79,
|
| 82 |
+
"l": 80,
|
| 83 |
+
"m": 81,
|
| 84 |
+
"n": 82,
|
| 85 |
+
"o": 83,
|
| 86 |
+
"p": 84,
|
| 87 |
+
"q": 85,
|
| 88 |
+
"r": 86,
|
| 89 |
+
"s": 87,
|
| 90 |
+
"t": 88,
|
| 91 |
+
"u": 89,
|
| 92 |
+
"v": 90,
|
| 93 |
+
"w": 91,
|
| 94 |
+
"x": 92,
|
| 95 |
+
"y": 93,
|
| 96 |
+
"z": 94,
|
| 97 |
+
"{": 95,
|
| 98 |
+
"|": 96,
|
| 99 |
+
"}": 97,
|
| 100 |
+
"~": 98,
|
| 101 |
+
"<-180>": 99,
|
| 102 |
+
"<-179>": 100,
|
| 103 |
+
"<-178>": 101,
|
| 104 |
+
"<-177>": 102,
|
| 105 |
+
"<-176>": 103,
|
| 106 |
+
"<-175>": 104,
|
| 107 |
+
"<-174>": 105,
|
| 108 |
+
"<-173>": 106,
|
| 109 |
+
"<-172>": 107,
|
| 110 |
+
"<-171>": 108,
|
| 111 |
+
"<-170>": 109,
|
| 112 |
+
"<-169>": 110,
|
| 113 |
+
"<-168>": 111,
|
| 114 |
+
"<-167>": 112,
|
| 115 |
+
"<-166>": 113,
|
| 116 |
+
"<-165>": 114,
|
| 117 |
+
"<-164>": 115,
|
| 118 |
+
"<-163>": 116,
|
| 119 |
+
"<-162>": 117,
|
| 120 |
+
"<-161>": 118,
|
| 121 |
+
"<-160>": 119,
|
| 122 |
+
"<-159>": 120,
|
| 123 |
+
"<-158>": 121,
|
| 124 |
+
"<-157>": 122,
|
| 125 |
+
"<-156>": 123,
|
| 126 |
+
"<-155>": 124,
|
| 127 |
+
"<-154>": 125,
|
| 128 |
+
"<-153>": 126,
|
| 129 |
+
"<-152>": 127,
|
| 130 |
+
"<-151>": 128,
|
| 131 |
+
"<-150>": 129,
|
| 132 |
+
"<-149>": 130,
|
| 133 |
+
"<-148>": 131,
|
| 134 |
+
"<-147>": 132,
|
| 135 |
+
"<-146>": 133,
|
| 136 |
+
"<-145>": 134,
|
| 137 |
+
"<-144>": 135,
|
| 138 |
+
"<-143>": 136,
|
| 139 |
+
"<-142>": 137,
|
| 140 |
+
"<-141>": 138,
|
| 141 |
+
"<-140>": 139,
|
| 142 |
+
"<-139>": 140,
|
| 143 |
+
"<-138>": 141,
|
| 144 |
+
"<-137>": 142,
|
| 145 |
+
"<-136>": 143,
|
| 146 |
+
"<-135>": 144,
|
| 147 |
+
"<-134>": 145,
|
| 148 |
+
"<-133>": 146,
|
| 149 |
+
"<-132>": 147,
|
| 150 |
+
"<-131>": 148,
|
| 151 |
+
"<-130>": 149,
|
| 152 |
+
"<-129>": 150,
|
| 153 |
+
"<-128>": 151,
|
| 154 |
+
"<-127>": 152,
|
| 155 |
+
"<-126>": 153,
|
| 156 |
+
"<-125>": 154,
|
| 157 |
+
"<-124>": 155,
|
| 158 |
+
"<-123>": 156,
|
| 159 |
+
"<-122>": 157,
|
| 160 |
+
"<-121>": 158,
|
| 161 |
+
"<-120>": 159,
|
| 162 |
+
"<-119>": 160,
|
| 163 |
+
"<-118>": 161,
|
| 164 |
+
"<-117>": 162,
|
| 165 |
+
"<-116>": 163,
|
| 166 |
+
"<-115>": 164,
|
| 167 |
+
"<-114>": 165,
|
| 168 |
+
"<-113>": 166,
|
| 169 |
+
"<-112>": 167,
|
| 170 |
+
"<-111>": 168,
|
| 171 |
+
"<-110>": 169,
|
| 172 |
+
"<-109>": 170,
|
| 173 |
+
"<-108>": 171,
|
| 174 |
+
"<-107>": 172,
|
| 175 |
+
"<-106>": 173,
|
| 176 |
+
"<-105>": 174,
|
| 177 |
+
"<-104>": 175,
|
| 178 |
+
"<-103>": 176,
|
| 179 |
+
"<-102>": 177,
|
| 180 |
+
"<-101>": 178,
|
| 181 |
+
"<-100>": 179,
|
| 182 |
+
"<-99>": 180,
|
| 183 |
+
"<-98>": 181,
|
| 184 |
+
"<-97>": 182,
|
| 185 |
+
"<-96>": 183,
|
| 186 |
+
"<-95>": 184,
|
| 187 |
+
"<-94>": 185,
|
| 188 |
+
"<-93>": 186,
|
| 189 |
+
"<-92>": 187,
|
| 190 |
+
"<-91>": 188,
|
| 191 |
+
"<-90>": 189,
|
| 192 |
+
"<-89>": 190,
|
| 193 |
+
"<-88>": 191,
|
| 194 |
+
"<-87>": 192,
|
| 195 |
+
"<-86>": 193,
|
| 196 |
+
"<-85>": 194,
|
| 197 |
+
"<-84>": 195,
|
| 198 |
+
"<-83>": 196,
|
| 199 |
+
"<-82>": 197,
|
| 200 |
+
"<-81>": 198,
|
| 201 |
+
"<-80>": 199,
|
| 202 |
+
"<-79>": 200,
|
| 203 |
+
"<-78>": 201,
|
| 204 |
+
"<-77>": 202,
|
| 205 |
+
"<-76>": 203,
|
| 206 |
+
"<-75>": 204,
|
| 207 |
+
"<-74>": 205,
|
| 208 |
+
"<-73>": 206,
|
| 209 |
+
"<-72>": 207,
|
| 210 |
+
"<-71>": 208,
|
| 211 |
+
"<-70>": 209,
|
| 212 |
+
"<-69>": 210,
|
| 213 |
+
"<-68>": 211,
|
| 214 |
+
"<-67>": 212,
|
| 215 |
+
"<-66>": 213,
|
| 216 |
+
"<-65>": 214,
|
| 217 |
+
"<-64>": 215,
|
| 218 |
+
"<-63>": 216,
|
| 219 |
+
"<-62>": 217,
|
| 220 |
+
"<-61>": 218,
|
| 221 |
+
"<-60>": 219,
|
| 222 |
+
"<-59>": 220,
|
| 223 |
+
"<-58>": 221,
|
| 224 |
+
"<-57>": 222,
|
| 225 |
+
"<-56>": 223,
|
| 226 |
+
"<-55>": 224,
|
| 227 |
+
"<-54>": 225,
|
| 228 |
+
"<-53>": 226,
|
| 229 |
+
"<-52>": 227,
|
| 230 |
+
"<-51>": 228,
|
| 231 |
+
"<-50>": 229,
|
| 232 |
+
"<-49>": 230,
|
| 233 |
+
"<-48>": 231,
|
| 234 |
+
"<-47>": 232,
|
| 235 |
+
"<-46>": 233,
|
| 236 |
+
"<-45>": 234,
|
| 237 |
+
"<-44>": 235,
|
| 238 |
+
"<-43>": 236,
|
| 239 |
+
"<-42>": 237,
|
| 240 |
+
"<-41>": 238,
|
| 241 |
+
"<-40>": 239,
|
| 242 |
+
"<-39>": 240,
|
| 243 |
+
"<-38>": 241,
|
| 244 |
+
"<-37>": 242,
|
| 245 |
+
"<-36>": 243,
|
| 246 |
+
"<-35>": 244,
|
| 247 |
+
"<-34>": 245,
|
| 248 |
+
"<-33>": 246,
|
| 249 |
+
"<-32>": 247,
|
| 250 |
+
"<-31>": 248,
|
| 251 |
+
"<-30>": 249,
|
| 252 |
+
"<-29>": 250,
|
| 253 |
+
"<-28>": 251,
|
| 254 |
+
"<-27>": 252,
|
| 255 |
+
"<-26>": 253,
|
| 256 |
+
"<-25>": 254,
|
| 257 |
+
"<-24>": 255,
|
| 258 |
+
"<-23>": 256,
|
| 259 |
+
"<-22>": 257,
|
| 260 |
+
"<-21>": 258,
|
| 261 |
+
"<-20>": 259,
|
| 262 |
+
"<-19>": 260,
|
| 263 |
+
"<-18>": 261,
|
| 264 |
+
"<-17>": 262,
|
| 265 |
+
"<-16>": 263,
|
| 266 |
+
"<-15>": 264,
|
| 267 |
+
"<-14>": 265,
|
| 268 |
+
"<-13>": 266,
|
| 269 |
+
"<-12>": 267,
|
| 270 |
+
"<-11>": 268,
|
| 271 |
+
"<-10>": 269,
|
| 272 |
+
"<-9>": 270,
|
| 273 |
+
"<-8>": 271,
|
| 274 |
+
"<-7>": 272,
|
| 275 |
+
"<-6>": 273,
|
| 276 |
+
"<-5>": 274,
|
| 277 |
+
"<-4>": 275,
|
| 278 |
+
"<-3>": 276,
|
| 279 |
+
"<-2>": 277,
|
| 280 |
+
"<-1>": 278,
|
| 281 |
+
"<0>": 279,
|
| 282 |
+
"<1>": 280,
|
| 283 |
+
"<2>": 281,
|
| 284 |
+
"<3>": 282,
|
| 285 |
+
"<4>": 283,
|
| 286 |
+
"<5>": 284,
|
| 287 |
+
"<6>": 285,
|
| 288 |
+
"<7>": 286,
|
| 289 |
+
"<8>": 287,
|
| 290 |
+
"<9>": 288,
|
| 291 |
+
"<10>": 289,
|
| 292 |
+
"<11>": 290,
|
| 293 |
+
"<12>": 291,
|
| 294 |
+
"<13>": 292,
|
| 295 |
+
"<14>": 293,
|
| 296 |
+
"<15>": 294,
|
| 297 |
+
"<16>": 295,
|
| 298 |
+
"<17>": 296,
|
| 299 |
+
"<18>": 297,
|
| 300 |
+
"<19>": 298,
|
| 301 |
+
"<20>": 299,
|
| 302 |
+
"<21>": 300,
|
| 303 |
+
"<22>": 301,
|
| 304 |
+
"<23>": 302,
|
| 305 |
+
"<24>": 303,
|
| 306 |
+
"<25>": 304,
|
| 307 |
+
"<26>": 305,
|
| 308 |
+
"<27>": 306,
|
| 309 |
+
"<28>": 307,
|
| 310 |
+
"<29>": 308,
|
| 311 |
+
"<30>": 309,
|
| 312 |
+
"<31>": 310,
|
| 313 |
+
"<32>": 311,
|
| 314 |
+
"<33>": 312,
|
| 315 |
+
"<34>": 313,
|
| 316 |
+
"<35>": 314,
|
| 317 |
+
"<36>": 315,
|
| 318 |
+
"<37>": 316,
|
| 319 |
+
"<38>": 317,
|
| 320 |
+
"<39>": 318,
|
| 321 |
+
"<40>": 319,
|
| 322 |
+
"<41>": 320,
|
| 323 |
+
"<42>": 321,
|
| 324 |
+
"<43>": 322,
|
| 325 |
+
"<44>": 323,
|
| 326 |
+
"<45>": 324,
|
| 327 |
+
"<46>": 325,
|
| 328 |
+
"<47>": 326,
|
| 329 |
+
"<48>": 327,
|
| 330 |
+
"<49>": 328,
|
| 331 |
+
"<50>": 329,
|
| 332 |
+
"<51>": 330,
|
| 333 |
+
"<52>": 331,
|
| 334 |
+
"<53>": 332,
|
| 335 |
+
"<54>": 333,
|
| 336 |
+
"<55>": 334,
|
| 337 |
+
"<56>": 335,
|
| 338 |
+
"<57>": 336,
|
| 339 |
+
"<58>": 337,
|
| 340 |
+
"<59>": 338,
|
| 341 |
+
"<60>": 339,
|
| 342 |
+
"<61>": 340,
|
| 343 |
+
"<62>": 341,
|
| 344 |
+
"<63>": 342,
|
| 345 |
+
"<64>": 343,
|
| 346 |
+
"<65>": 344,
|
| 347 |
+
"<66>": 345,
|
| 348 |
+
"<67>": 346,
|
| 349 |
+
"<68>": 347,
|
| 350 |
+
"<69>": 348,
|
| 351 |
+
"<70>": 349,
|
| 352 |
+
"<71>": 350,
|
| 353 |
+
"<72>": 351,
|
| 354 |
+
"<73>": 352,
|
| 355 |
+
"<74>": 353,
|
| 356 |
+
"<75>": 354,
|
| 357 |
+
"<76>": 355,
|
| 358 |
+
"<77>": 356,
|
| 359 |
+
"<78>": 357,
|
| 360 |
+
"<79>": 358,
|
| 361 |
+
"<80>": 359,
|
| 362 |
+
"<81>": 360,
|
| 363 |
+
"<82>": 361,
|
| 364 |
+
"<83>": 362,
|
| 365 |
+
"<84>": 363,
|
| 366 |
+
"<85>": 364,
|
| 367 |
+
"<86>": 365,
|
| 368 |
+
"<87>": 366,
|
| 369 |
+
"<88>": 367,
|
| 370 |
+
"<89>": 368,
|
| 371 |
+
"<90>": 369,
|
| 372 |
+
"<91>": 370,
|
| 373 |
+
"<92>": 371,
|
| 374 |
+
"<93>": 372,
|
| 375 |
+
"<94>": 373,
|
| 376 |
+
"<95>": 374,
|
| 377 |
+
"<96>": 375,
|
| 378 |
+
"<97>": 376,
|
| 379 |
+
"<98>": 377,
|
| 380 |
+
"<99>": 378,
|
| 381 |
+
"<100>": 379,
|
| 382 |
+
"<101>": 380,
|
| 383 |
+
"<102>": 381,
|
| 384 |
+
"<103>": 382,
|
| 385 |
+
"<104>": 383,
|
| 386 |
+
"<105>": 384,
|
| 387 |
+
"<106>": 385,
|
| 388 |
+
"<107>": 386,
|
| 389 |
+
"<108>": 387,
|
| 390 |
+
"<109>": 388,
|
| 391 |
+
"<110>": 389,
|
| 392 |
+
"<111>": 390,
|
| 393 |
+
"<112>": 391,
|
| 394 |
+
"<113>": 392,
|
| 395 |
+
"<114>": 393,
|
| 396 |
+
"<115>": 394,
|
| 397 |
+
"<116>": 395,
|
| 398 |
+
"<117>": 396,
|
| 399 |
+
"<118>": 397,
|
| 400 |
+
"<119>": 398,
|
| 401 |
+
"<120>": 399,
|
| 402 |
+
"<121>": 400,
|
| 403 |
+
"<122>": 401,
|
| 404 |
+
"<123>": 402,
|
| 405 |
+
"<124>": 403,
|
| 406 |
+
"<125>": 404,
|
| 407 |
+
"<126>": 405,
|
| 408 |
+
"<127>": 406,
|
| 409 |
+
"<128>": 407,
|
| 410 |
+
"<129>": 408,
|
| 411 |
+
"<130>": 409,
|
| 412 |
+
"<131>": 410,
|
| 413 |
+
"<132>": 411,
|
| 414 |
+
"<133>": 412,
|
| 415 |
+
"<134>": 413,
|
| 416 |
+
"<135>": 414,
|
| 417 |
+
"<136>": 415,
|
| 418 |
+
"<137>": 416,
|
| 419 |
+
"<138>": 417,
|
| 420 |
+
"<139>": 418,
|
| 421 |
+
"<140>": 419,
|
| 422 |
+
"<141>": 420,
|
| 423 |
+
"<142>": 421,
|
| 424 |
+
"<143>": 422,
|
| 425 |
+
"<144>": 423,
|
| 426 |
+
"<145>": 424,
|
| 427 |
+
"<146>": 425,
|
| 428 |
+
"<147>": 426,
|
| 429 |
+
"<148>": 427,
|
| 430 |
+
"<149>": 428,
|
| 431 |
+
"<150>": 429,
|
| 432 |
+
"<151>": 430,
|
| 433 |
+
"<152>": 431,
|
| 434 |
+
"<153>": 432,
|
| 435 |
+
"<154>": 433,
|
| 436 |
+
"<155>": 434,
|
| 437 |
+
"<156>": 435,
|
| 438 |
+
"<157>": 436,
|
| 439 |
+
"<158>": 437,
|
| 440 |
+
"<159>": 438,
|
| 441 |
+
"<160>": 439,
|
| 442 |
+
"<161>": 440,
|
| 443 |
+
"<162>": 441,
|
| 444 |
+
"<163>": 442,
|
| 445 |
+
"<164>": 443,
|
| 446 |
+
"<165>": 444,
|
| 447 |
+
"<166>": 445,
|
| 448 |
+
"<167>": 446,
|
| 449 |
+
"<168>": 447,
|
| 450 |
+
"<169>": 448,
|
| 451 |
+
"<170>": 449,
|
| 452 |
+
"<171>": 450,
|
| 453 |
+
"<172>": 451,
|
| 454 |
+
"<173>": 452,
|
| 455 |
+
"<174>": 453,
|
| 456 |
+
"<175>": 454,
|
| 457 |
+
"<176>": 455,
|
| 458 |
+
"<177>": 456,
|
| 459 |
+
"<178>": 457,
|
| 460 |
+
"<179>": 458,
|
| 461 |
+
"<180>": 459,
|
| 462 |
+
"<std>": 460,
|
| 463 |
+
"<aug>": 461,
|
| 464 |
+
"<unk>": 462
|
| 465 |
+
}
|
trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eaa57b21440d492010fce4b33ef28aaad8bdc8a300eb9c9d8a3fe5a889e6a302
|
| 3 |
+
size 5560
|