Update main.py
Browse files
main.py
CHANGED
@@ -1,21 +1,19 @@
|
|
1 |
from fastapi import FastAPI
|
2 |
from pydantic import BaseModel
|
3 |
-
from llama_cpp import Llama
|
4 |
import uvicorn
|
5 |
import prompt_style
|
6 |
import os
|
7 |
-
from huggingface_hub import
|
8 |
|
9 |
|
10 |
-
model_id = "failspy/Meta-Llama-3-8B-Instruct-abliterated-v3
|
11 |
-
|
12 |
-
model = Llama(model_path=model_path, n_gpu_layers=-1, n_ctx=4096, verbose=False)
|
13 |
|
14 |
class Item(BaseModel):
|
15 |
prompt: str
|
16 |
history: list
|
17 |
system_prompt: str
|
18 |
-
temperature: float = 0.
|
19 |
max_new_tokens: int = 1024
|
20 |
top_p: float = 0.95
|
21 |
repetition_penalty: float = 1.0
|
@@ -34,14 +32,35 @@ def format_prompt(item: Item):
|
|
34 |
return messages
|
35 |
|
36 |
def generate(item: Item):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
formatted_prompt = format_prompt(item)
|
38 |
-
|
39 |
-
|
40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
|
42 |
|
43 |
-
out = output['choices'][0]['message']['content']
|
44 |
-
return out
|
45 |
|
46 |
@app.post("/generate/")
|
47 |
async def generate_text(item: Item):
|
|
|
1 |
from fastapi import FastAPI
|
2 |
from pydantic import BaseModel
|
|
|
3 |
import uvicorn
|
4 |
import prompt_style
|
5 |
import os
|
6 |
+
from huggingface_hub import InferenceClient
|
7 |
|
8 |
|
9 |
+
model_id = "failspy/Meta-Llama-3-8B-Instruct-abliterated-v3"
|
10 |
+
client = InferenceClient(model_id, token=os.environ['HF_TOKEN'])
|
|
|
11 |
|
12 |
class Item(BaseModel):
|
13 |
prompt: str
|
14 |
history: list
|
15 |
system_prompt: str
|
16 |
+
temperature: float = 0.8
|
17 |
max_new_tokens: int = 1024
|
18 |
top_p: float = 0.95
|
19 |
repetition_penalty: float = 1.0
|
|
|
32 |
return messages
|
33 |
|
34 |
def generate(item: Item):
|
35 |
+
temperature = float(item.temperature)
|
36 |
+
if temperature < 1e-2:
|
37 |
+
temperature = 1e-2
|
38 |
+
top_p = float(item.top_p)
|
39 |
+
|
40 |
+
generate_kwargs = dict(
|
41 |
+
temperature=temperature,
|
42 |
+
max_new_tokens=item.max_new_tokens,
|
43 |
+
top_p=top_p,
|
44 |
+
repetition_penalty=item.repetition_penalty,
|
45 |
+
do_sample=True,
|
46 |
+
seed=item.seed,
|
47 |
+
)
|
48 |
+
|
49 |
formatted_prompt = format_prompt(item)
|
50 |
+
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
51 |
+
output = ""
|
52 |
+
|
53 |
+
for response in stream:
|
54 |
+
output += response.token.text
|
55 |
+
return output
|
56 |
+
|
57 |
+
# output = model.create_chat_completion(messages=formatted_prompt, seed=item.seed,
|
58 |
+
# temperature=item.temperature,
|
59 |
+
# max_tokens=item.max_new_tokens)
|
60 |
|
61 |
|
62 |
+
# out = output['choices'][0]['message']['content']
|
63 |
+
# return out
|
64 |
|
65 |
@app.post("/generate/")
|
66 |
async def generate_text(item: Item):
|