bart-large/config.json

65 lines
1.4 KiB
JSON
Raw Normal View History

2020-02-12 19:53:37 +00:00
{
"activation_dropout": 0.0,
2020-02-13 21:46:16 +00:00
"architectures": [
"BartModel",
"BartForMaskedLM",
"BartForSequenceClassification"
],
2020-02-12 19:53:37 +00:00
"attention_dropout": 0.0,
"classif_dropout": 0.0,
"d_model": 1024,
"decoder_attention_heads": 16,
"decoder_ffn_dim": 4096,
"decoder_layerdrop": 0.0,
"decoder_layers": 12,
2020-04-02 11:13:57 +00:00
"decoder_start_token_id": 2,
2020-02-12 19:53:37 +00:00
"do_sample": false,
"dropout": 0.1,
"encoder_attention_heads": 16,
"encoder_ffn_dim": 4096,
"encoder_layerdrop": 0.0,
"encoder_layers": 12,
"finetuning_task": null,
"id2label": {
"0": "LABEL_0",
2020-02-13 22:29:14 +00:00
"1": "LABEL_1",
"2": "LABEL_2"
2020-02-12 19:53:37 +00:00
},
"init_std": 0.02,
"is_decoder": false,
"label2id": {
"LABEL_0": 0,
2020-02-13 22:29:14 +00:00
"LABEL_1": 1,
"LABEL_2": 2
2020-02-12 19:53:37 +00:00
},
"length_penalty": 1.0,
"max_length": 20,
"max_position_embeddings": 1024,
"model_type": "bart",
"num_beams": 1,
"num_hidden_layers": 12,
2020-02-13 22:29:14 +00:00
"num_labels": 3,
2020-02-12 19:53:37 +00:00
"num_return_sequences": 1,
"output_attentions": false,
"output_hidden_states": false,
2020-02-13 21:46:16 +00:00
"output_past": false,
2020-03-30 15:01:09 +00:00
"prefix": " ",
2020-02-12 19:53:37 +00:00
"pruned_heads": {},
"repetition_penalty": 1.0,
2020-04-02 11:13:57 +00:00
"task_specific_params": {
"summarization": {
"early_stopping": true,
"length_penalty": 2.0,
"max_length": 142,
"min_length": 56,
"no_repeat_ngram_size": 3,
"num_beams": 4
}
},
2020-02-12 19:53:37 +00:00
"temperature": 1.0,
"top_k": 50,
"top_p": 1.0,
"torchscript": false,
"use_bfloat16": false,
"vocab_size": 50265
2020-03-30 15:01:09 +00:00
}