Upload tokenizer.chat_template
Browse files- tokenizer.chat_template +31 -0
tokenizer.chat_template
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
import torch
|
3 |
+
from transformers import AutoTokenizer,AutoModelForMaskedLM, GPT2LMHeadModel,GPT2Tokenizer, AutoModelForSeq2SeqLM, AutoModelForCausalLM
|
4 |
+
tokenizer = GPT2Tokenizer.from_pretrained("microsoft/DialoGPT-medium")
|
5 |
+
|
6 |
+
model = GPT2LMHeadModel.from_pretrained('Stage v3.0')
|
7 |
+
# Let's chat for 4 lines
|
8 |
+
|
9 |
+
for step in range(50):
|
10 |
+
# encode the new user input, add the eos_token and return a tensor in Pytorch
|
11 |
+
new_user_input_ids = tokenizer.encode(input(">> You:") + tokenizer.eos_token, return_tensors='pt')
|
12 |
+
# print(new_user_input_ids)
|
13 |
+
|
14 |
+
|
15 |
+
# append the new user input tokens to the chat history
|
16 |
+
bot_input_ids = torch.cat([ new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
|
17 |
+
|
18 |
+
# generated a response while limiting the total chat history to 1000 tokens,
|
19 |
+
chat_history_ids = model.generate(
|
20 |
+
bot_input_ids, max_length=200,
|
21 |
+
pad_token_id=tokenizer.eos_token_id,
|
22 |
+
no_repeat_ngram_size=3,
|
23 |
+
do_sample=True,
|
24 |
+
top_k=100, # It controls the diversity of the generated output; the model considers the top 100 tokens
|
25 |
+
top_p=0.9,# tokens with a cumulative probability higher than 0.9 are excluded.
|
26 |
+
temperature=0.9 # It controls the randomness of the generated output
|
27 |
+
)
|
28 |
+
|
29 |
+
# pretty print last ouput tokens from bot
|
30 |
+
print("Chatbot: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))
|
31 |
+
|