tomaarsen HF Staff commited on
Commit
0ae81dc
·
verified ·
1 Parent(s): 03de46e

Update config.json

Browse files
Files changed (1) hide show
  1. config.json +41 -41
config.json CHANGED
@@ -1,41 +1,41 @@
1
- {
2
- "architectures": [
3
- "EuroBertForMaskedLM"
4
- ],
5
- "attention_bias": false,
6
- "attention_dropout": 0.0,
7
- "bos_token": "<|begin_of_text|>",
8
- "bos_token_id": 128000,
9
- "clf_pooling": "late",
10
- "dtype": "float32",
11
- "eos_token": "<|end_of_text|>",
12
- "eos_token_id": 128001,
13
- "head_dim": 64,
14
- "hidden_act": "silu",
15
- "hidden_dropout": 0.0,
16
- "hidden_size": 32,
17
- "initializer_range": 0.02,
18
- "intermediate_size": 64,
19
- "is_causal": false,
20
- "is_decoder": false,
21
- "mask_token": "<|mask|>",
22
- "mask_token_id": 128002,
23
- "max_position_embeddings": 8192,
24
- "mlp_bias": false,
25
- "model_type": "eurobert",
26
- "num_attention_heads": 2,
27
- "num_hidden_layers": 4,
28
- "num_key_value_heads": 2,
29
- "pad_token": "<|end_of_text|>",
30
- "pad_token_id": 128001,
31
- "pretraining_tp": 1,
32
- "rms_norm_eps": 1e-05,
33
- "rope_parameters": {
34
- "rope_theta": 250000,
35
- "rope_type": "default"
36
- },
37
- "tie_word_embeddings": false,
38
- "transformers_version": "5.3.0.dev0",
39
- "use_cache": false,
40
- "vocab_size": 128256
41
- }
 
1
+ {
2
+ "architectures": [
3
+ "EuroBertForMaskedLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token": "<|begin_of_text|>",
8
+ "bos_token_id": 128000,
9
+ "classifier_pooling": "late",
10
+ "dtype": "float32",
11
+ "eos_token": "<|end_of_text|>",
12
+ "eos_token_id": 128001,
13
+ "head_dim": 64,
14
+ "hidden_act": "silu",
15
+ "hidden_dropout": 0.0,
16
+ "hidden_size": 32,
17
+ "initializer_range": 0.02,
18
+ "intermediate_size": 64,
19
+ "is_causal": false,
20
+ "is_decoder": false,
21
+ "mask_token": "<|mask|>",
22
+ "mask_token_id": 128002,
23
+ "max_position_embeddings": 8192,
24
+ "mlp_bias": false,
25
+ "model_type": "eurobert",
26
+ "num_attention_heads": 2,
27
+ "num_hidden_layers": 4,
28
+ "num_key_value_heads": 2,
29
+ "pad_token": "<|end_of_text|>",
30
+ "pad_token_id": 128001,
31
+ "pretraining_tp": 1,
32
+ "rms_norm_eps": 1e-05,
33
+ "rope_parameters": {
34
+ "rope_theta": 250000,
35
+ "rope_type": "default"
36
+ },
37
+ "tie_word_embeddings": false,
38
+ "transformers_version": "5.3.0.dev0",
39
+ "use_cache": false,
40
+ "vocab_size": 128256
41
+ }