Shaelois commited on
Commit
ca7720c
·
verified ·
1 Parent(s): 1357461

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +44 -43
config.json CHANGED
@@ -1,43 +1,44 @@
1
- {
2
- "_name_or_path": "results/checkpoint-4026",
3
- "activation_dropout": 0.0,
4
- "activation_function": "gelu_new",
5
- "architectures": [
6
- "BigBirdPegasusForConditionalGeneration"
7
- ],
8
- "attention_dropout": 0.0,
9
- "attention_type": "block_sparse",
10
- "block_size": 64,
11
- "bos_token_id": 2,
12
- "classifier_dropout": 0.0,
13
- "d_model": 1024,
14
- "decoder_attention_heads": 16,
15
- "decoder_ffn_dim": 4096,
16
- "decoder_layerdrop": 0.0,
17
- "decoder_layers": 16,
18
- "decoder_start_token_id": 2,
19
- "dropout": 0.1,
20
- "encoder_attention_heads": 16,
21
- "encoder_ffn_dim": 4096,
22
- "encoder_layerdrop": 0.0,
23
- "encoder_layers": 16,
24
- "eos_token_id": 1,
25
- "gradient_checkpointing": false,
26
- "init_std": 0.02,
27
- "is_encoder_decoder": true,
28
- "length_penalty": null,
29
- "max_length": null,
30
- "max_position_embeddings": 4096,
31
- "model_type": "bigbird_pegasus",
32
- "num_beams": null,
33
- "num_hidden_layers": 16,
34
- "num_random_blocks": 3,
35
- "pad_token_id": 0,
36
- "scale_embedding": true,
37
- "tokenizer_class": "PegasusTokenizer",
38
- "torch_dtype": "float32",
39
- "transformers_version": "4.49.0",
40
- "use_bias": false,
41
- "use_cache": true,
42
- "vocab_size": 96103
43
- }
 
 
1
+ {
2
+ "_attn_implementation_autoset": true,
3
+ "_name_or_path": "results\\checkpoint-4026",
4
+ "activation_dropout": 0.0,
5
+ "activation_function": "gelu_new",
6
+ "architectures": [
7
+ "BigBirdPegasusForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "attention_type": "block_sparse",
11
+ "block_size": 64,
12
+ "bos_token_id": 2,
13
+ "classifier_dropout": 0.0,
14
+ "d_model": 1024,
15
+ "decoder_attention_heads": 16,
16
+ "decoder_ffn_dim": 4096,
17
+ "decoder_layerdrop": 0.0,
18
+ "decoder_layers": 16,
19
+ "decoder_start_token_id": 2,
20
+ "dropout": 0.1,
21
+ "encoder_attention_heads": 16,
22
+ "encoder_ffn_dim": 4096,
23
+ "encoder_layerdrop": 0.0,
24
+ "encoder_layers": 16,
25
+ "eos_token_id": 1,
26
+ "gradient_checkpointing": false,
27
+ "init_std": 0.02,
28
+ "is_encoder_decoder": true,
29
+ "length_penalty": null,
30
+ "max_length": null,
31
+ "max_position_embeddings": 4096,
32
+ "model_type": "bigbird_pegasus",
33
+ "num_beams": null,
34
+ "num_hidden_layers": 16,
35
+ "num_random_blocks": 3,
36
+ "pad_token_id": 0,
37
+ "scale_embedding": true,
38
+ "tokenizer_class": "PegasusTokenizer",
39
+ "torch_dtype": "float32",
40
+ "transformers_version": "4.49.0",
41
+ "use_bias": false,
42
+ "use_cache": true,
43
+ "vocab_size": 96103
44
+ }