Update model.py
Browse files
model.py
CHANGED
@@ -0,0 +1,125 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-nc-2.0
|
3 |
+
datasets:
|
4 |
+
- wikimedia/wikipedia
|
5 |
+
- open-llm-leaderboard-old/details_microsoft__DialoGPT-large
|
6 |
+
- li2017dailydialog/daily_dialog
|
7 |
+
- google/Synthetic-Persona-Chat
|
8 |
+
- kanhatakeyama/CommonCrawl-RAG-QA-Calm3-22b-chat
|
9 |
+
- yuyijiong/Multi-doc-QA-CommonCrawl
|
10 |
+
- Skylion007/openwebtext
|
11 |
+
- Bingsu/openwebtext_20p
|
12 |
+
- segyges/OpenWebText2
|
13 |
+
language:
|
14 |
+
- en
|
15 |
+
library_name: transformers
|
16 |
+
---
|
17 |
+
import torch
|
18 |
+
from transformers import BartForConditionalGeneration, BartTokenizer, GPT2LMHeadModel, GPT2Tokenizer
|
19 |
+
import argparse
|
20 |
+
import sys
|
21 |
+
|
22 |
+
class AdvancedSummarizer:
|
23 |
+
def __init__(self, model_name="facebook/bart-large-cnn"):
|
24 |
+
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
25 |
+
self.model = BartForConditionalGeneration.from_pretrained(model_name).to(self.device)
|
26 |
+
self.tokenizer = BartTokenizer.from_pretrained(model_name)
|
27 |
+
|
28 |
+
def summarize(self, text, max_length=150, min_length=50, length_penalty=2.0, num_beams=4):
|
29 |
+
inputs = self.tokenizer([text], max_length=1024, return_tensors="pt", truncation=True)
|
30 |
+
inputs = inputs.to(self.device)
|
31 |
+
|
32 |
+
summary_ids = self.model.generate(
|
33 |
+
inputs["input_ids"],
|
34 |
+
num_beams=num_beams,
|
35 |
+
max_length=max_length,
|
36 |
+
min_length=min_length,
|
37 |
+
length_penalty=length_penalty
|
38 |
+
)
|
39 |
+
|
40 |
+
summary = self.tokenizer.decode(summary_ids[0], skip_special_tokens=True)
|
41 |
+
return summary
|
42 |
+
|
43 |
+
def main_summarizer():
|
44 |
+
# Example usage
|
45 |
+
summarizer = AdvancedSummarizer()
|
46 |
+
text = """...""" # Your text here
|
47 |
+
summary = summarizer.summarize(text)
|
48 |
+
print("Summary:")
|
49 |
+
print(summary)
|
50 |
+
|
51 |
+
class AdvancedTextGenerator:
|
52 |
+
def __init__(self, model_name="gpt2-medium"):
|
53 |
+
try:
|
54 |
+
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
55 |
+
print(f"Using device: {self.device}")
|
56 |
+
self.model = GPT2LMHeadModel.from_pretrained(model_name).to(self.device)
|
57 |
+
self.tokenizer = GPT2Tokenizer.from_pretrained(model_name)
|
58 |
+
except Exception as e:
|
59 |
+
print(f"Error initializing the model: {e}")
|
60 |
+
sys.exit(1)
|
61 |
+
|
62 |
+
def generate_text(self, prompt, max_length=100, num_return_sequences=1,
|
63 |
+
temperature=1.0, top_k=50, top_p=0.95, repetition_penalty=1.0):
|
64 |
+
try:
|
65 |
+
input_ids = self.tokenizer.encode(prompt, return_tensors="pt").to(self.device)
|
66 |
+
|
67 |
+
output_sequences = self.model.generate(
|
68 |
+
input_ids=input_ids,
|
69 |
+
max_length=max_length + len(input_ids[0]),
|
70 |
+
temperature=temperature,
|
71 |
+
top_k=top_k,
|
72 |
+
top_p=top_p,
|
73 |
+
repetition_penalty=repetition_penalty,
|
74 |
+
do_sample=True,
|
75 |
+
num_return_sequences=num_return_sequences,
|
76 |
+
)
|
77 |
+
|
78 |
+
generated_sequences = []
|
79 |
+
for generated_sequence in output_sequences:
|
80 |
+
text = self.tokenizer.decode(generated_sequence, clean_up_tokenization_spaces=True)
|
81 |
+
total_sequence = text[len(self.tokenizer.decode(input_ids[0], clean_up_tokenization_spaces=True)):]
|
82 |
+
generated_sequences.append(total_sequence)
|
83 |
+
|
84 |
+
return generated_sequences
|
85 |
+
except Exception as e:
|
86 |
+
return [f"Error during text generation: {e}"]
|
87 |
+
|
88 |
+
def main_generator():
|
89 |
+
parser = argparse.ArgumentParser(description="Advanced Text Generator")
|
90 |
+
parser.add_argument("--prompt", type=str, help="Starting prompt for text generation")
|
91 |
+
parser.add_argument("--max_length", type=int, default=100, help="Maximum length of generated text")
|
92 |
+
parser.add_argument("--num_sequences", type=int, default=1, help="Number of sequences to generate")
|
93 |
+
parser.add_argument("--temperature", type=float, default=1.0, help="Temperature for sampling")
|
94 |
+
parser.add_argument("--top_k", type=int, default=50, help="Top-k sampling parameter")
|
95 |
+
parser.add_argument("--top_p", type=float, default=0.95, help="Top-p sampling parameter")
|
96 |
+
parser.add_argument("--repetition_penalty", type=float, default=1.0, help="Repetition penalty")
|
97 |
+
|
98 |
+
args = parser.parse_args()
|
99 |
+
|
100 |
+
generator = AdvancedTextGenerator()
|
101 |
+
|
102 |
+
if args.prompt:
|
103 |
+
prompt = args.prompt
|
104 |
+
else:
|
105 |
+
print("Please enter the prompt for text generation:")
|
106 |
+
prompt = input().strip()
|
107 |
+
|
108 |
+
generated_texts = generator.generate_text(
|
109 |
+
prompt,
|
110 |
+
max_length=args.max_length,
|
111 |
+
num_return_sequences=args.num_sequences,
|
112 |
+
temperature=args.temperature,
|
113 |
+
top_k=args.top_k,
|
114 |
+
top_p=args.top_p,
|
115 |
+
repetition_penalty=args.repetition_penalty
|
116 |
+
)
|
117 |
+
|
118 |
+
print("\nGenerated Text(s):")
|
119 |
+
for i, text in enumerate(generated_texts, 1):
|
120 |
+
print(f"\n--- Sequence {i} ---")
|
121 |
+
print(text)
|
122 |
+
|
123 |
+
if __name__ == "__main__":
|
124 |
+
main_summarizer() # Call the summarizer main function
|
125 |
+
main_generator() # Call the text generator main function
|