File size: 1,755 Bytes
37b0e2d 377986e 37b0e2d b4751ce 65f5921 37b0e2d 65f5921 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 |
import torch
from typing import Any, Dict
from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig
class EndpointHandler:
def __init__(self, path=""):
with torch.autocast('cuda'):
# load model and tokenizer from path
self.tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neox-20b", padding_side="left")
config = AutoConfig.from_pretrained(path, trust_remote_code=True)
# config.attn_config['attn_impl'] = 'triton'
config.init_device = 'cuda:0' # For fast initialization directly on GPU!
config.max_seq_len = 4096 # (input + output) tokens can now be up to 4096
self.model = AutoModelForCausalLM.from_pretrained(
path,
config,
torch_dtype=torch.float16,
trust_remote_code=True
)
# self.device = "cuda" if torch.cuda.is_available() else "cpu"
self.device = 'cuda'
def __call__(self, data: Dict[str, Any]) -> Dict[str, str]:
# process input
inputs = data.pop("inputs", data)
parameters = data.pop("parameters", None)
with torch.autocast('cuda'):
# preprocess
inputs = self.tokenizer(inputs, return_tensors="pt").to(self.device)
# pass inputs with all kwargs in data
if parameters is not None:
outputs = self.model.generate(**inputs, **parameters)
else:
outputs = self.model.generate(**inputs)
# postprocess the prediction
prediction = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
return [{"generated_text": prediction}] |