njwright92
commited on
Update handler.py
Browse files- handler.py +59 -27
handler.py
CHANGED
@@ -1,25 +1,42 @@
|
|
1 |
from typing import Dict, List, Any
|
2 |
from llama_cpp import Llama
|
3 |
import gemma_tools
|
|
|
4 |
|
5 |
MAX_TOKENS = 1000
|
6 |
|
7 |
|
8 |
-
class EndpointHandler
|
9 |
-
def __init__(self, model_dir=None):
|
10 |
-
|
11 |
-
|
12 |
-
# Initialize the Llama model directly
|
13 |
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
|
20 |
def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
|
21 |
-
|
|
|
22 |
|
|
|
|
|
|
|
|
|
23 |
args_check = gemma_tools.get_args_or_none(data)
|
24 |
|
25 |
if not args_check[0]: # If validation failed
|
@@ -29,26 +46,25 @@ class EndpointHandler():
|
|
29 |
"description": args_check.get("description", "Validation error in arguments")
|
30 |
}]
|
31 |
|
32 |
-
|
|
|
33 |
|
34 |
-
# Define the formatting template
|
35 |
-
|
36 |
|
37 |
try:
|
38 |
-
formatted_prompt =
|
39 |
-
|
40 |
except Exception as e:
|
41 |
-
|
42 |
return [{
|
43 |
"status": "error",
|
44 |
"reason": "Invalid format",
|
45 |
"detail": str(e)
|
46 |
}]
|
47 |
|
|
|
48 |
max_length = data.get("max_length", 212)
|
49 |
try:
|
50 |
max_length = int(max_length)
|
51 |
-
|
52 |
except ValueError:
|
53 |
return [{
|
54 |
"status": "error",
|
@@ -56,16 +72,32 @@ class EndpointHandler():
|
|
56 |
"detail": "max_length was not a valid integer"
|
57 |
}]
|
58 |
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
|
67 |
return [{
|
68 |
"status": "success",
|
69 |
-
#
|
70 |
-
"response": res['choices'][0]['text']
|
71 |
}]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
from typing import Dict, List, Any
|
2 |
from llama_cpp import Llama
|
3 |
import gemma_tools
|
4 |
+
import os
|
5 |
|
6 |
MAX_TOKENS = 1000
|
7 |
|
8 |
|
9 |
+
class EndpointHandler:
|
10 |
+
def __init__(self, model_dir: str = None):
|
11 |
+
"""
|
12 |
+
Initialize the EndpointHandler with the path to the model directory.
|
|
|
13 |
|
14 |
+
:param model_dir: Path to the directory containing the model file.
|
15 |
+
"""
|
16 |
+
if model_dir:
|
17 |
+
# Update the model filename to match the one in your repository
|
18 |
+
model_path = os.path.join(
|
19 |
+
model_dir, "comic_mistral-v5.2.q5_0.gguf")
|
20 |
+
if not os.path.exists(model_path):
|
21 |
+
raise FileNotFoundError(
|
22 |
+
f"The model file was not found at {model_path}")
|
23 |
+
|
24 |
+
try:
|
25 |
+
self.model = Llama(
|
26 |
+
model_path=model_path,
|
27 |
+
n_ctx=MAX_TOKENS, # Use n_ctx for context size in llama_cpp
|
28 |
+
)
|
29 |
+
except Exception as e:
|
30 |
+
raise RuntimeError(f"Failed to load the model: {e}")
|
31 |
|
32 |
def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
|
33 |
+
"""
|
34 |
+
Handle incoming requests for model inference.
|
35 |
|
36 |
+
:param data: Dictionary containing input data and parameters for the model.
|
37 |
+
:return: A list with a dictionary containing the status and response or error details.
|
38 |
+
"""
|
39 |
+
# Extract and validate arguments from the data
|
40 |
args_check = gemma_tools.get_args_or_none(data)
|
41 |
|
42 |
if not args_check[0]: # If validation failed
|
|
|
46 |
"description": args_check.get("description", "Validation error in arguments")
|
47 |
}]
|
48 |
|
49 |
+
# If validation passed, args are in the second element of the tuple
|
50 |
+
args = args_check[1]
|
51 |
|
52 |
+
# Define the formatting template for the prompt
|
53 |
+
prompt_format = "<startofturn>system\n{system_prompt} <endofturn>\n<startofturn>user\n{inputs} <endofturn>\n<startofturn>model"
|
54 |
|
55 |
try:
|
56 |
+
formatted_prompt = prompt_format.format(**args)
|
|
|
57 |
except Exception as e:
|
|
|
58 |
return [{
|
59 |
"status": "error",
|
60 |
"reason": "Invalid format",
|
61 |
"detail": str(e)
|
62 |
}]
|
63 |
|
64 |
+
# Parse max_length, default to 212 if not provided or invalid
|
65 |
max_length = data.get("max_length", 212)
|
66 |
try:
|
67 |
max_length = int(max_length)
|
|
|
68 |
except ValueError:
|
69 |
return [{
|
70 |
"status": "error",
|
|
|
72 |
"detail": "max_length was not a valid integer"
|
73 |
}]
|
74 |
|
75 |
+
# Perform inference
|
76 |
+
try:
|
77 |
+
res = self.model(
|
78 |
+
formatted_prompt,
|
79 |
+
temperature=args["temperature"],
|
80 |
+
top_p=args["top_p"],
|
81 |
+
top_k=args["top_k"],
|
82 |
+
max_tokens=max_length
|
83 |
+
)
|
84 |
+
except Exception as e:
|
85 |
+
return [{
|
86 |
+
"status": "error",
|
87 |
+
"reason": "Inference failed",
|
88 |
+
"detail": str(e)
|
89 |
+
}]
|
90 |
|
91 |
return [{
|
92 |
"status": "success",
|
93 |
+
# Extract the text from the response
|
94 |
+
"response": res['choices'][0]['text'].strip()
|
95 |
}]
|
96 |
+
|
97 |
+
|
98 |
+
# Usage in your script or where the handler is instantiated:
|
99 |
+
try:
|
100 |
+
handler = EndpointHandler("/repository")
|
101 |
+
except (FileNotFoundError, RuntimeError) as e:
|
102 |
+
print(f"Initialization error: {e}")
|
103 |
+
exit(1) # Exit with an error code if the handler cannot be initialized
|