Create handler.py
#18
by
wensun
- opened
- handler.py +23 -0
- requirements.txt +6 -0
handler.py
ADDED
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import Dict, List, Any
|
2 |
+
from transformers import AutoTokenizer, AutoModel
|
3 |
+
import torch
|
4 |
+
|
5 |
+
class EndpointHandler:
|
6 |
+
def __init__(self, path=""):
|
7 |
+
# load model and processor from path
|
8 |
+
self.tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True)
|
9 |
+
self.model = AutoModel.from_pretrained(path, trust_remote_code=True).half().cuda()
|
10 |
+
|
11 |
+
def __call__(self, data: Dict[str, Any]) -> Dict[str, str]:
|
12 |
+
"""
|
13 |
+
Args:
|
14 |
+
data (:dict:):
|
15 |
+
The payload with the text prompt and generation parameters.
|
16 |
+
"""
|
17 |
+
# process input
|
18 |
+
inputs = data.pop("inputs", data)
|
19 |
+
history = data.pop("history", None)
|
20 |
+
|
21 |
+
response, new_history = self.model.chat(self.tokenizer, inputs, history)
|
22 |
+
|
23 |
+
return [{"generated_text": response, "history": new_history}]
|
requirements.txt
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
protobuf>=3.19.5,<3.20.1
|
2 |
+
transformers==4.26.1
|
3 |
+
icetk
|
4 |
+
cpm_kernels
|
5 |
+
torch>=1.10
|
6 |
+
gradio
|