Update inference.py
Browse files- inference.py +18 -8
inference.py
CHANGED
@@ -1,7 +1,8 @@
|
|
1 |
# inference.py
|
2 |
-
|
3 |
-
from typing import List, Dict, Generator, Optional
|
4 |
from hf_client import get_inference_client
|
|
|
|
|
5 |
|
6 |
def chat_completion(
|
7 |
model_id: str,
|
@@ -14,20 +15,25 @@ def chat_completion(
|
|
14 |
|
15 |
Args:
|
16 |
model_id: The model identifier to use.
|
17 |
-
messages: A list of OpenAI
|
18 |
provider: Optional override for provider; uses model default if None.
|
19 |
max_tokens: Maximum tokens to generate.
|
20 |
|
21 |
Returns:
|
22 |
The assistant's response content.
|
23 |
"""
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
26 |
model=model_id,
|
27 |
messages=messages,
|
28 |
max_tokens=max_tokens
|
29 |
)
|
30 |
-
return
|
31 |
|
32 |
|
33 |
def stream_chat_completion(
|
@@ -35,12 +41,16 @@ def stream_chat_completion(
|
|
35 |
messages: List[Dict[str, str]],
|
36 |
provider: Optional[str] = None,
|
37 |
max_tokens: int = 4096
|
38 |
-
)
|
39 |
"""
|
40 |
Generator for streaming chat completions.
|
41 |
Yields partial message chunks as strings.
|
42 |
"""
|
43 |
-
|
|
|
|
|
|
|
|
|
44 |
stream = client.chat.completions.create(
|
45 |
model=model_id,
|
46 |
messages=messages,
|
|
|
1 |
# inference.py
|
2 |
+
from typing import List, Dict, Optional
|
|
|
3 |
from hf_client import get_inference_client
|
4 |
+
from models import find_model
|
5 |
+
|
6 |
|
7 |
def chat_completion(
|
8 |
model_id: str,
|
|
|
15 |
|
16 |
Args:
|
17 |
model_id: The model identifier to use.
|
18 |
+
messages: A list of OpenAI-style {'role','content'} messages.
|
19 |
provider: Optional override for provider; uses model default if None.
|
20 |
max_tokens: Maximum tokens to generate.
|
21 |
|
22 |
Returns:
|
23 |
The assistant's response content.
|
24 |
"""
|
25 |
+
# resolve default provider from registry if needed
|
26 |
+
if provider is None:
|
27 |
+
meta = find_model(model_id)
|
28 |
+
provider = meta.default_provider if meta else "auto"
|
29 |
+
|
30 |
+
client = get_inference_client(model_id, provider)
|
31 |
+
resp = client.chat.completions.create(
|
32 |
model=model_id,
|
33 |
messages=messages,
|
34 |
max_tokens=max_tokens
|
35 |
)
|
36 |
+
return resp.choices[0].message.content
|
37 |
|
38 |
|
39 |
def stream_chat_completion(
|
|
|
41 |
messages: List[Dict[str, str]],
|
42 |
provider: Optional[str] = None,
|
43 |
max_tokens: int = 4096
|
44 |
+
):
|
45 |
"""
|
46 |
Generator for streaming chat completions.
|
47 |
Yields partial message chunks as strings.
|
48 |
"""
|
49 |
+
if provider is None:
|
50 |
+
meta = find_model(model_id)
|
51 |
+
provider = meta.default_provider if meta else "auto"
|
52 |
+
|
53 |
+
client = get_inference_client(model_id, provider)
|
54 |
stream = client.chat.completions.create(
|
55 |
model=model_id,
|
56 |
messages=messages,
|