language:
- en
thumbnail: null
tags:
- gpt2
- conversational
license: apache-2.0
datasets:
- wikipedia-turkish
metrics:
- perplexity
- accuracy
widget:
- text: Bu yazıyı bir bilgisayar yazdı. Yazarken
context: ''
- text: İnternete kolay erişim sayesinde dünya daha da küçüldü. Bunun sonucunda
context: ''
GPT2 Persona Chatbot based on Movie Characters
Model used for https://www.metayazar.com/chatbot GPT2 Small Trained on movie scripts (especially Sci-fi) This work is based on Persona Chatbot originally done by Hugging Face team (https://medium.com/huggingface/how-to-build-a-state-of-the-art-conversational-ai-with-transfer-learning-2d818ac26313) For cleaning movie scripts I also provide cleaner code https://github.com/gorkemgoknar/moviescriptcleaner
Example persona how to: https://gist.github.com/gorkemgoknar/ae29bf9d14fa814e6a64d0e57a4a4ed7
For obvious reasons I cannot share raw personafile but you can check above gist for example how to create it.
A working "full" demo can be seen in https://www.metayazar.com/chatbot For Turkish version (with limited training) https://www.metayazar.com/chatbot_tr
tokenizer = AutoTokenizer.from_pretrained('microsoft/DialoGPT-small')
model = AutoModelWithLMHead.from_pretrained('output-small')
# Let's chat for 5 lines
for step in range(100):
# encode the new user input, add the eos_token and return a tensor in Pytorch
new_user_input_ids = tokenizer.encode(input(">> User:") + tokenizer.eos_token, return_tensors='pt')
# print(new_user_input_ids)
# append the new user input tokens to the chat history
bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
# generated a response while limiting the total chat history to 1000 tokens,
chat_history_ids = model.generate(
bot_input_ids, max_length=500,
pad_token_id=tokenizer.eos_token_id,
no_repeat_ngram_size=3,
do_sample=True,
top_k=100,
top_p=0.7,
temperature = 0.8
)
# pretty print last ouput tokens from bot
print("AI: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))