Page MenuHomec4science
Files F91212782

tok.py
No OneTemporary
Actions

File Metadata

Created
Sat, Nov 9, 00:56

tok.py
View Options

import os
import json
from tokenizers import ByteLevelBPETokenizer
from transformers import GPT2Tokenizer
def train_tok(gpt_path, vocab_size):
tokenizer = ByteLevelBPETokenizer()
os.chdir(gpt_path)
print("train tokenizer ...")
tokenizer.train(files="train.txt", vocab_size=vocab_size, min_frequency=2,
special_tokens=["<input>", "<pad>", "<unk>", "<find>", "<answer>", "<|endoftext|>"])
print("... done training")
# save tokenizer
if not os.path.exists('tok'):
os.makedirs('tok')
tokenizer.save_model("tok")
# get the path of the tokenizer files
os.chdir('tok')
tok_path = os.getcwd()
# print(tok_path)
os.chdir(gpt_path)
# create the configuration files
config = {
"_num_labels": 2,
"activation_function": "gelu_new",
"architectures": [
"GPT2LMHeadModel"
],
"attn_pdrop": 0.1,
"do_sample": False,
"early_stopping": False,
"embd_pdrop": 0.1,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"is_decoder": False,
"is_encoder_decoder": False,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"layer_norm_epsilon": 1e-05,
"length_penalty": 1.0,
"max_length": 20,
"min_length": 0,
"model_type": "gpt2",
"n_ctx": 1024,
"n_embd": 768,
"n_head": 12,
"n_layer": 12,
"n_positions": 1024,
"no_repeat_ngram_size": 0,
"num_beams": 1,
"num_return_sequences": 1,
"output_attentions": False,
"output_hidden_states": False,
"output_past": True,
"pruned_heads": {},
"repetition_penalty": 1.0,
"resid_pdrop": 0.1,
"summary_first_dropout": 0.1,
"summary_proj_to_labels": True,
"summary_type": "cls_index",
"summary_use_proj": True,
"temperature": 1.0,
"top_k": 50,
"top_p": 1.0,
"torchscript": False,
"use_bfloat16": False,
"vocab_size": 50257
}
print("save config")
with open(tok_path + "/config.json", 'w') as fp:
json.dump(config, fp)
tokenizer_config = {
"max_len": 1024
}
with open(tok_path + "/tokenizer_config.json", 'w') as fp:
json.dump(tokenizer_config, fp)
return tok_path
def get_tok(tok_path):
tokenizer = GPT2Tokenizer.from_pretrained(tok_path)
return tokenizer

Event Timeline

c4science · Help