Commit
•
41f3c42
1
Parent(s):
97dd7a1
Delete Chatbot.md
Browse files- Chatbot.md +0 -104
Chatbot.md
DELETED
@@ -1,104 +0,0 @@
|
|
1 |
-
## Requirements
|
2 |
-
Be sure to install necessary libraries.
|
3 |
-
|
4 |
-
```python
|
5 |
-
pip install torch
|
6 |
-
pip install transformers
|
7 |
-
pip install nltk
|
8 |
-
pip install pyspellchecker
|
9 |
-
```
|
10 |
-
|
11 |
-
## Engaging with VergilGPT2
|
12 |
-
If you're eager to have a conversation with VergilGPT2, you can utilize the following code snippet. Feel free to experiment with the variables 'temperature', 'top_k', 'top_p', to customize the response generation according to your preferences.
|
13 |
-
|
14 |
-
```python
|
15 |
-
import torch
|
16 |
-
import re
|
17 |
-
import random
|
18 |
-
import nltk
|
19 |
-
from transformers import AutoTokenizer, AutoModelForCausalLM
|
20 |
-
from nltk.corpus import words
|
21 |
-
import string
|
22 |
-
from spellchecker import SpellChecker
|
23 |
-
|
24 |
-
spell = SpellChecker(language='en') # specify the language as English
|
25 |
-
|
26 |
-
def filter_english_words(text):
|
27 |
-
# Tokenize the input text
|
28 |
-
tokenized_text = nltk.word_tokenize(text)
|
29 |
-
# Initialize an empty list to store the filtered words
|
30 |
-
filtered_text = []
|
31 |
-
# Iterate through the tokenized text
|
32 |
-
for word in tokenized_text:
|
33 |
-
# Check if the word is in the English words set or is a punctuation
|
34 |
-
if word.lower() in spell or word in string.punctuation:
|
35 |
-
# If it is, append it to the filtered text list
|
36 |
-
filtered_text.append(word)
|
37 |
-
# Return the filtered text as a string
|
38 |
-
return ' '.join(filtered_text)
|
39 |
-
|
40 |
-
def generate_response(model, tokenizer, input_text, max_length=300, min_length=20, num_return_sequences=1, temperature=0.2, top_k=50, top_p=0.9, num_beams=10, repetition_penalty=1.0):
|
41 |
-
input_ids = tokenizer.encode(input_text, return_tensors='pt')
|
42 |
-
|
43 |
-
# Set model to eval mode
|
44 |
-
model.eval()
|
45 |
-
|
46 |
-
# Generate responses using different decoding strategies
|
47 |
-
try:
|
48 |
-
output = model.generate(
|
49 |
-
input_ids=input_ids,
|
50 |
-
max_length=max_length,
|
51 |
-
min_length=min_length,
|
52 |
-
num_return_sequences=num_return_sequences,
|
53 |
-
temperature=temperature,
|
54 |
-
top_k=top_k,
|
55 |
-
top_p=top_p,
|
56 |
-
num_beams=num_beams,
|
57 |
-
no_repeat_ngram_size=2,
|
58 |
-
do_sample=True,
|
59 |
-
repetition_penalty=repetition_penalty # Added repetition_penalty
|
60 |
-
)
|
61 |
-
|
62 |
-
# Decode the generated responses
|
63 |
-
responses = [tokenizer.decode(o, skip_special_tokens=True) for o in output]
|
64 |
-
# Remove input_text from the responses
|
65 |
-
responses = [response[len(tokenizer.decode(input_ids[0], skip_special_tokens=True)):] for response in responses]
|
66 |
-
|
67 |
-
# Filter out non-English words
|
68 |
-
responses = [filter_english_words(response) for response in responses]
|
69 |
-
|
70 |
-
return responses
|
71 |
-
|
72 |
-
except RuntimeError:
|
73 |
-
return ["I'm sorry, I encountered an error while generating a response."]
|
74 |
-
|
75 |
-
|
76 |
-
# Load pre-trained model and tokenizer
|
77 |
-
access_token = "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"
|
78 |
-
model_id = "Starcodium/VergilGPT2"
|
79 |
-
tokenizer = AutoTokenizer.from_pretrained(model_id, revision="main", use_auth_token=access_token)
|
80 |
-
model = AutoModelForCausalLM.from_pretrained(model_id, revision="main", use_auth_token=access_token)
|
81 |
-
|
82 |
-
tokenizer.pad_token = tokenizer.eos_token
|
83 |
-
model.config.pad_token_id = model.config.eos_token_id
|
84 |
-
|
85 |
-
# Get user input and generate responses
|
86 |
-
while True:
|
87 |
-
input_text = input("Type 'quit' or 'exit' to stop runtime.\nEnter your input text: ")
|
88 |
-
if input_text.lower() in ["quit", "exit"]:
|
89 |
-
break
|
90 |
-
|
91 |
-
responses = generate_response(model, tokenizer, input_text, max_length=100, min_length=20, num_return_sequences=1, temperature=0.7, top_k=40, top_p=0.5, num_beams=1, repetition_penalty=1.2)
|
92 |
-
responses = [r for r in responses if r.strip() != '']
|
93 |
-
if responses:
|
94 |
-
response = responses[0]
|
95 |
-
else:
|
96 |
-
response = "I'm sorry, I don't have a response at the moment."
|
97 |
-
|
98 |
-
# Print the bot's response
|
99 |
-
print("Vergil: "+response)
|
100 |
-
```
|
101 |
-
|
102 |
-
This code snippet allows you to engage in conversations with VergilGPT2. Simply enter your input text, and VergilGPT2 will generate responses based on the provided context. Experiment with different values of the variables temperature, top_k, and top_p to customize the response generation process according to your desired preferences.
|
103 |
-
|
104 |
-
Please note that the code assumes you have access to the Starcodium/VergilGPT2 model and its associated tokenizer. Ensure you have the required authentication token (access_token) to access the model and tokenizer.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|