Update README.md
Browse files
README.md
CHANGED
@@ -5,4 +5,45 @@ language:
|
|
5 |
- en
|
6 |
---
|
7 |
|
8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
- en
|
6 |
---
|
7 |
|
8 |
+
# HomerBot: A conversational chatbot imitating Homer Simpson
|
9 |
+
|
10 |
+
This model is a fine-tuned [DialoGPT](https://huggingface.co/microsoft/DialoGPT-medium) (medium version) on Simpsons [scripts](https://www.kaggle.com/datasets/pierremegret/dialogue-lines-of-the-simpsons).
|
11 |
+
|
12 |
+
More specifically, we fine-tune DialoGPT-medium for 3 epochs on 10K **(character utterance, Homer's response)** pairs
|
13 |
+
|
14 |
+
For more details, check out our git [repo](https://github.com/jesseDingley/HomerBot) containing all the code.
|
15 |
+
|
16 |
+
### How to use
|
17 |
+
|
18 |
+
|
19 |
+
```python
|
20 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
21 |
+
import torch
|
22 |
+
|
23 |
+
tokenizer = AutoTokenizer.from_pretrained("DingleyMaillotUrgell/homer-bot")
|
24 |
+
model = AutoModelForCausalLM.from_pretrained("DingleyMaillotUrgell/homer-bot")
|
25 |
+
|
26 |
+
# Let's chat for 5 lines
|
27 |
+
for step in range(5):
|
28 |
+
# encode the new user input, add the eos_token and return a tensor in Pytorch
|
29 |
+
new_user_input_ids = tokenizer.encode(input(">> User: ") + tokenizer.eos_token, return_tensors='pt')
|
30 |
+
|
31 |
+
# append the new user input tokens to the chat history
|
32 |
+
bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
|
33 |
+
|
34 |
+
# generated a response while limiting the total chat history to 1000 tokens,
|
35 |
+
chat_history_ids = model.generate(
|
36 |
+
bot_input_ids,
|
37 |
+
max_length=1000,
|
38 |
+
pad_token_id=tokenizer.eos_token_id,
|
39 |
+
no_repeat_ngram_size=3,
|
40 |
+
do_sample=True,
|
41 |
+
top_k=100,
|
42 |
+
top_p=0.7,
|
43 |
+
temperature = 0.8
|
44 |
+
)
|
45 |
+
|
46 |
+
# print last outpput tokens from bot
|
47 |
+
print("Homer: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))
|
48 |
+
```
|
49 |
+
|