Advertisement
Guest User

Untitled

a guest
May 19th, 2019
107
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 0.77 KB | None | 0 0
  1. # use this for BERT-base
  2.  
  3. bert_base_config = {
  4. "attention_probs_dropout_prob": 0.1,
  5. "directionality": "bidi",
  6. "hidden_act": "gelu",
  7. "hidden_dropout_prob": 0.1,
  8. "hidden_size": 768,
  9. "initializer_range": 0.02,
  10. "intermediate_size": 3072,
  11. "max_position_embeddings": 512,
  12. "num_attention_heads": 12,
  13. "num_hidden_layers": 12,
  14. "pooler_fc_size": 768,
  15. "pooler_num_attention_heads": 12,
  16. "pooler_num_fc_layers": 3,
  17. "pooler_size_per_head": 128,
  18. "pooler_type": "first_token_transform",
  19. "type_vocab_size": 2,
  20. "vocab_size": VOC_SIZE
  21. }
  22.  
  23. with open("{}/bert_config.json".format(MODEL_DIR), "w") as fo:
  24. json.dump(bert_base_config, fo, indent=2)
  25.  
  26. with open("{}/{}".format(MODEL_DIR, VOC_FNAME), "w") as fo:
  27. for token in bert_vocab:
  28. fo.write(token+"\n")
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement