alvinfadli commited on
Commit
9ff3ea3
·
verified ·
1 Parent(s): ef6db40

Upload 2 files

Browse files
Files changed (2) hide show
  1. main.py +25 -0
  2. requirements.txt +6 -0
main.py ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from ctransformers import AutoModelForCausalLM
2
+ from fastapi import FastAPI
3
+ from pydantic import BaseModel
4
+
5
+
6
+ llm = AutoModelForCausalLM.from_pretrained("llama2-tuned.gguf",
7
+ model_type='llama-2',
8
+ max_new_tokens = 1096,
9
+ threads = 3,
10
+ )
11
+
12
+ #Pydantic object
13
+ class validation(BaseModel):
14
+ prompt: str
15
+ #Fast API
16
+ app = FastAPI()
17
+
18
+ @app.post("/llm_on_cpu")
19
+ async def stream(item: validation):
20
+ system_prompt = 'Below is an instruction that describes a task. Write a response that appropriately completes the request.'
21
+ prompt = f"""<s>[INST] <<SYS>>
22
+ {system_prompt}
23
+ <</SYS>>
24
+ {item.prompt} [/INST]"""
25
+ return llm(prompt)
requirements.txt ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ python-multipart
2
+ fastapi
3
+ pydantic
4
+ uvicorn
5
+ requests
6
+ ctransformers