Update app.py
Browse files
app.py
CHANGED
@@ -1,18 +1,29 @@
|
|
1 |
import torch
|
2 |
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
|
3 |
|
|
|
4 |
model_name = "deepseek-ai/deepseek-math-7b-instruct"
|
5 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
6 |
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")
|
7 |
model.generation_config = GenerationConfig.from_pretrained(model_name)
|
8 |
model.generation_config.pad_token_id = model.generation_config.eos_token_id
|
9 |
|
10 |
-
|
11 |
-
|
12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
input_tensor = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
|
|
|
|
|
14 |
outputs = model.generate(input_tensor.to(model.device), max_new_tokens=100)
|
15 |
|
|
|
16 |
result = tokenizer.decode(outputs[0][input_tensor.shape[1]:], skip_special_tokens=True)
|
17 |
print(result)
|
18 |
-
|
|
|
1 |
import torch
|
2 |
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
|
3 |
|
4 |
+
# Specify the model and tokenizer
|
5 |
model_name = "deepseek-ai/deepseek-math-7b-instruct"
|
6 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
7 |
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")
|
8 |
model.generation_config = GenerationConfig.from_pretrained(model_name)
|
9 |
model.generation_config.pad_token_id = model.generation_config.eos_token_id
|
10 |
|
11 |
+
# Function to read text from a file
|
12 |
+
def read_input_text(file_path):
|
13 |
+
with open(file_path, 'r', encoding='utf-8') as file:
|
14 |
+
text = file.read()
|
15 |
+
return text.strip()
|
16 |
+
|
17 |
+
# Example usage: Replace 'input.txt' with your file path
|
18 |
+
input_text = read_input_text('input.txt')
|
19 |
+
|
20 |
+
# Prepare input as a chat message
|
21 |
+
messages = [{"role": "user", "content": input_text}]
|
22 |
input_tensor = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
|
23 |
+
|
24 |
+
# Generate outputs from the model
|
25 |
outputs = model.generate(input_tensor.to(model.device), max_new_tokens=100)
|
26 |
|
27 |
+
# Decode the generated output
|
28 |
result = tokenizer.decode(outputs[0][input_tensor.shape[1]:], skip_special_tokens=True)
|
29 |
print(result)
|
|