Create handler.py
Browse files- handler.py +60 -0
handler.py
ADDED
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig, pipeline
|
2 |
+
from transformers_stream_generator import init_stream_support
|
3 |
+
init_stream_support()
|
4 |
+
|
5 |
+
template = """Alice Gate's Persona: Alice Gate is a young, computer engineer-nerd with a knack for problem solving and a passion for technology.
|
6 |
+
<START>
|
7 |
+
{user_name}: So how did you get into computer engineering?
|
8 |
+
Alice Gate: I've always loved tinkering with technology since I was a kid.
|
9 |
+
{user_name}: That's really impressive!
|
10 |
+
Alice Gate: *She chuckles bashfully* Thanks!
|
11 |
+
{user_name}: So what do you do when you're not working on computers?
|
12 |
+
Alice Gate: I love exploring, going out with friends, watching movies, and playing video games.
|
13 |
+
{user_name}: What's your favorite type of computer hardware to work with?
|
14 |
+
Alice Gate: Motherboards, they're like puzzles and the backbone of any system.
|
15 |
+
{user_name}: That sounds great!
|
16 |
+
Alice Gate: Yeah, it's really fun. I'm lucky to be able to do this as a job.
|
17 |
+
<END>
|
18 |
+
Alice Gate: *Alice strides into the room with a smile, her eyes lighting up when she sees you. She's wearing a light blue t-shirt and jeans, her laptop bag slung over one shoulder. She takes a seat next to you, her enthusiasm palpable in the air* Hey! I'm so excited to finally meet you. I've heard so many great things about you and I'm eager to pick your brain about computers. I'm sure you have a wealth of knowledge that I can learn from. *She grins, eyes twinkling with excitement* Let's get started!
|
19 |
+
"""
|
20 |
+
|
21 |
+
class EndpointHandler():
|
22 |
+
|
23 |
+
def __init__(self, path=""):
|
24 |
+
quantization_config = BitsAndBytesConfig(
|
25 |
+
load_in_8bit = True,
|
26 |
+
llm_int8_threshold = 0.0,
|
27 |
+
llm_int8_enable_fp32_cpu_offload = True
|
28 |
+
)
|
29 |
+
self.tokenizer = AutoTokenizer.from_pretrained(path)
|
30 |
+
self.model = AutoModelForCausalLM.from_pretrained(
|
31 |
+
path,
|
32 |
+
device_map = "auto"
|
33 |
+
torch_dtype = "auto",
|
34 |
+
low_cpu_mem_usage = True,
|
35 |
+
quantization_config = quantization_config
|
36 |
+
)
|
37 |
+
|
38 |
+
def __call__(self, data):
|
39 |
+
prompt += data.pop("inputs", data)
|
40 |
+
input_ids = self.tokenizer(
|
41 |
+
prompt,
|
42 |
+
return_tensors="pt"
|
43 |
+
) .input_ids
|
44 |
+
stream_generator = self.model.generate(
|
45 |
+
input_ids,
|
46 |
+
max_new_tokens = 70,
|
47 |
+
do_sample = True,
|
48 |
+
do_stream = True,
|
49 |
+
temperature = 0.5,
|
50 |
+
top_p = 0.9,
|
51 |
+
top_k = 0,
|
52 |
+
repetition_penalty = 1.1,
|
53 |
+
pad_token_id = 50256,
|
54 |
+
num_return_sequences = 1
|
55 |
+
)
|
56 |
+
result = []
|
57 |
+
for token in stream_generator:
|
58 |
+
result.append(self.tokenizer.decode(token))
|
59 |
+
if result[-1] == "\n":
|
60 |
+
return "".join(result).strip()
|