deepparag commited on
Commit
d00a411
1 Parent(s): 6a27ae3

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +61 -0
README.md ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ thumbnail: https://cdn.discordapp.com/app-icons/870239976690970625/c02cae78ae105f07969cfd8f8ea3d0a0.png
3
+ tags:
4
+ - conversational
5
+ license: mit
6
+ ---
7
+ An generative AI made using [microsoft/DialoGPT-small](https://huggingface.co/microsoft/DialoGPT-small).
8
+
9
+ Trained on:
10
+
11
+ https://www.kaggle.com/Cornell-University/movie-dialog-corpus
12
+
13
+ https://www.kaggle.com/jef1056/discord-data
14
+
15
+
16
+
17
+ Important:
18
+
19
+ The AI can be a bit weird at times as it is still undergoing training!
20
+
21
+ At times it send stuff using :<random_wierd_words>: as they are discord emotes.
22
+
23
+ It also send random @RandomName as it is trying to ping people.
24
+
25
+ This works well on discord but on the web not so much but it is easy enough to remove such stuff using [re.sub](https://docs.python.org/3/library/re.html#re.sub)
26
+
27
+
28
+
29
+ Issues:
30
+
31
+ The AI like with all conversation AI lacks a character, it changes its name way too often. This can be solved using an AIML chatbot to give it a stable character!
32
+
33
+ [Live Demo](https://dumbot-331213.uc.r.appspot.com/)
34
+
35
+ Example:
36
+ ```python
37
+ from transformers import AutoTokenizer, AutoModelWithLMHead
38
+
39
+ tokenizer = AutoTokenizer.from_pretrained("deepparag/DumBot")
40
+ model = AutoModelWithLMHead.from_pretrained("deepparag/DumBot")
41
+ # Let's chat for 4 lines
42
+ for step in range(4):
43
+ # encode the new user input, add the eos_token and return a tensor in Pytorch
44
+ new_user_input_ids = tokenizer.encode(input(">> User:") + tokenizer.eos_token, return_tensors='pt')
45
+ # print(new_user_input_ids)
46
+ # append the new user input tokens to the chat history
47
+ bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids
48
+ # generated a response while limiting the total chat history to 1000 tokens,
49
+ chat_history_ids = model.generate(
50
+ bot_input_ids, max_length=200,
51
+ pad_token_id=tokenizer.eos_token_id,
52
+ no_repeat_ngram_size=4,
53
+ do_sample=True,
54
+ top_k=100,
55
+ top_p=0.7,
56
+ temperature=0.8
57
+ )
58
+
59
+ # pretty print last ouput tokens from bot
60
+ print("DumBot: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))
61
+ ```