Spaces:
Running
Running
Update hf_model.py
Browse files- hf_model.py +67 -31
hf_model.py
CHANGED
@@ -1,58 +1,94 @@
|
|
1 |
-
|
2 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
3 |
import torch
|
4 |
from datetime import datetime
|
5 |
import os
|
|
|
6 |
|
7 |
class HFModel:
|
8 |
def __init__(self, model_name):
|
9 |
parts = model_name.split("/")
|
10 |
self.friendly_name = parts[1]
|
11 |
-
self.model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True, torch_dtype=torch.bfloat16).cuda()
|
12 |
-
self.tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
|
13 |
self.chat_history = []
|
14 |
self.log_file = f"chat_log_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md"
|
15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
def generate_response(self, input_text, max_length=100, skip_special_tokens=True):
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
|
|
|
|
|
|
|
|
|
|
21 |
|
22 |
def stream_response(self, input_text, max_length=100, skip_special_tokens=True):
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
|
|
|
|
|
|
|
|
27 |
|
28 |
def chat(self, user_input, max_length=100, skip_special_tokens=True):
|
29 |
-
|
30 |
-
|
|
|
|
|
31 |
|
32 |
-
|
33 |
-
|
34 |
|
35 |
-
|
36 |
-
|
|
|
37 |
|
38 |
-
|
39 |
-
|
40 |
|
41 |
-
|
|
|
|
|
|
|
42 |
|
43 |
def save_chat_log(self):
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
|
|
|
|
49 |
|
50 |
def clear_chat_history(self):
|
51 |
-
|
52 |
-
|
|
|
|
|
|
|
|
|
53 |
|
54 |
def print_chat_history(self):
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
2 |
import torch
|
3 |
from datetime import datetime
|
4 |
import os
|
5 |
+
from MewUtilities.mew_log import log_info, log_error # Import your custom logging methods
|
6 |
|
7 |
class HFModel:
|
8 |
def __init__(self, model_name):
|
9 |
parts = model_name.split("/")
|
10 |
self.friendly_name = parts[1]
|
|
|
|
|
11 |
self.chat_history = []
|
12 |
self.log_file = f"chat_log_{datetime.now().strftime('%Y%m%d_%H%M%S')}.md"
|
13 |
|
14 |
+
try:
|
15 |
+
log_info(f"=== Loading Model: {self.friendly_name} ===")
|
16 |
+
self.model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True, torch_dtype=torch.bfloat16).cuda()
|
17 |
+
self.tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
|
18 |
+
log_info(f"=== Model Loaded Successfully: {self.friendly_name} ===")
|
19 |
+
except Exception as e:
|
20 |
+
log_error(f"ERROR Loading Model: {e}")
|
21 |
+
raise
|
22 |
+
|
23 |
def generate_response(self, input_text, max_length=100, skip_special_tokens=True):
|
24 |
+
try:
|
25 |
+
inputs = self.tokenizer(input_text, return_tensors="pt").to(self.model.device)
|
26 |
+
outputs = self.model.generate(**inputs, max_length=max_length)
|
27 |
+
response = self.tokenizer.decode(outputs[0], skip_special_tokens=skip_special_tokens).strip()
|
28 |
+
log_info(f"Generated Response: {response}")
|
29 |
+
return response
|
30 |
+
except Exception as e:
|
31 |
+
log_error(f"ERROR Generating Response: {e}")
|
32 |
+
raise
|
33 |
|
34 |
def stream_response(self, input_text, max_length=100, skip_special_tokens=True):
|
35 |
+
try:
|
36 |
+
inputs = self.tokenizer(input_text, return_tensors="pt").to(self.model.device)
|
37 |
+
for output in self.model.generate(**inputs, max_length=max_length, do_stream=True):
|
38 |
+
response = self.tokenizer.decode(output, skip_special_tokens=skip_special_tokens).strip()
|
39 |
+
yield response
|
40 |
+
except Exception as e:
|
41 |
+
log_error(f"ERROR Streaming Response: {e}")
|
42 |
+
raise
|
43 |
|
44 |
def chat(self, user_input, max_length=100, skip_special_tokens=True):
|
45 |
+
try:
|
46 |
+
# Add user input to chat history
|
47 |
+
self.chat_history.append({"role": "user", "content": user_input})
|
48 |
+
log_info(f"User Input: {user_input}")
|
49 |
|
50 |
+
# Generate model response
|
51 |
+
model_response = self.generate_response(user_input, max_length=max_length, skip_special_tokens=skip_special_tokens)
|
52 |
|
53 |
+
# Add model response to chat history
|
54 |
+
self.chat_history.append({"role": "assistant", "content": model_response})
|
55 |
+
log_info(f"Assistant Response: {model_response}")
|
56 |
|
57 |
+
# Save chat log
|
58 |
+
self.save_chat_log()
|
59 |
|
60 |
+
return model_response
|
61 |
+
except Exception as e:
|
62 |
+
log_error(f"ERROR in Chat: {e}")
|
63 |
+
raise
|
64 |
|
65 |
def save_chat_log(self):
|
66 |
+
try:
|
67 |
+
with open(self.log_file, "a", encoding="utf-8") as f:
|
68 |
+
for entry in self.chat_history[-2:]: # Save only the latest interaction
|
69 |
+
role = entry["role"]
|
70 |
+
content = entry["content"]
|
71 |
+
f.write(f"**{role.capitalize()}:**\n\n{content}\n\n---\n\n")
|
72 |
+
log_info(f"Chat log saved to {self.log_file}")
|
73 |
+
except Exception as e:
|
74 |
+
log_error(f"ERROR Saving Chat Log: {e}")
|
75 |
+
raise
|
76 |
|
77 |
def clear_chat_history(self):
|
78 |
+
try:
|
79 |
+
self.chat_history = []
|
80 |
+
log_info("Chat history cleared.")
|
81 |
+
except Exception as e:
|
82 |
+
log_error(f"ERROR Clearing Chat History: {e}")
|
83 |
+
raise
|
84 |
|
85 |
def print_chat_history(self):
|
86 |
+
try:
|
87 |
+
for entry in self.chat_history:
|
88 |
+
role = entry["role"]
|
89 |
+
content = entry["content"]
|
90 |
+
print(f"{role.capitalize()}: {content}\n")
|
91 |
+
log_info("Printed chat history to console.")
|
92 |
+
except Exception as e:
|
93 |
+
log_error(f"ERROR Printing Chat History: {e}")
|
94 |
+
raise
|