47 lines
1.7 KiB
Python
47 lines
1.7 KiB
Python
import torch
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
|
from typing import Dict, Any
|
|
|
|
class EndpointHandler:
|
|
def __init__(self, path=""):
|
|
quantization_config = BitsAndBytesConfig(
|
|
load_in_4bit=True,
|
|
bnb_4bit_compute_dtype=torch.float16,
|
|
bnb_4bit_quant_type="nf4",
|
|
)
|
|
self.tokenizer = AutoTokenizer.from_pretrained(path)
|
|
self.model = AutoModelForCausalLM.from_pretrained(
|
|
path,
|
|
quantization_config=quantization_config,
|
|
device_map="auto",
|
|
torch_dtype=torch.float16,
|
|
)
|
|
|
|
def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
|
|
inputs = data.get("inputs", "")
|
|
parameters = data.get("parameters", {})
|
|
|
|
if isinstance(inputs, list):
|
|
text = self.tokenizer.apply_chat_template(inputs, tokenize=False, add_generation_prompt=True)
|
|
else:
|
|
text = inputs
|
|
|
|
encoded = self.tokenizer(text, return_tensors="pt").to(self.model.device)
|
|
# Remove token_type_ids - not used by Mistral models
|
|
encoded.pop("token_type_ids", None)
|
|
|
|
max_new_tokens = parameters.get("max_new_tokens", 512)
|
|
temperature = parameters.get("temperature", 0.7)
|
|
|
|
with torch.no_grad():
|
|
outputs = self.model.generate(
|
|
**encoded,
|
|
max_new_tokens=max_new_tokens,
|
|
temperature=max(temperature, 0.01),
|
|
do_sample=temperature > 0,
|
|
)
|
|
|
|
new_tokens = outputs[0][encoded["input_ids"].shape[1]:]
|
|
response = self.tokenizer.decode(new_tokens, skip_special_tokens=True)
|
|
return [{"generated_text": response}]
|