aieng-lab/codet5p-220m_comment-type-python
010
1{2 "_name_or_path": "Salesforce/codet5p-220m",3 "architectures": [4 "T5ForSequenceClassification"5 ],6 "bos_token_id": 1,7 "classifier_dropout": 0.0,8 "d_ff": 3072,9 "d_kv": 64,10 "d_model": 768,11 "decoder_start_token_id": 0,12 "dense_act_fn": "relu",13 "dropout_rate": 0.1,14 "eos_token_id": 2,15 "feed_forward_proj": "relu",16 "id2label": {17 "0": "LABEL_0",18 "1": "LABEL_1",19 "2": "LABEL_2",20 "3": "LABEL_3",21 "4": "LABEL_4"22 },23 "initializer_factor": 1.0,24 "is_encoder_decoder": true,25 "is_gated_act": false,26 "label2id": {27 "LABEL_0": 0,28 "LABEL_1": 1,29 "LABEL_2": 2,30 "LABEL_3": 3,31 "LABEL_4": 432 },33 "layer_norm_epsilon": 1e-06,34 "model_type": "t5",35 "n_positions": 512,36 "num_decoder_layers": 12,37 "num_heads": 12,38 "num_layers": 12,39 "output_past": true,40 "pad_token_id": 0,41 "problem_type": "multi_label_classification",42 "relative_attention_max_distance": 128,43 "relative_attention_num_buckets": 32,44 "torch_dtype": "bfloat16",45 "transformers_version": "4.48.3",46 "use_cache": true,47 "vocab_size": 3210048}49 