Spaces:
Sleeping
Sleeping
File size: 1,519 Bytes
ff9f17d 98e48f2 ff9f17d |
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 46 47 48 49 50 51 52 53 54 55 56 |
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
import warnings
torch_dtype = torch.bfloat16
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model_name = "mosaicml/mpt-7b"
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch_dtype,
trust_remote_code=True,
use_auth_token=None,
)
tokenizer = AutoTokenizer.from_pretrained(
model_name,
trust_remote_code=True,
use_auth_token=None,
)
model.eval()
model.to(device=device, dtype=torch_dtype)
if tokenizer.pad_token_id is None:
warnings.warn(
"pad_token_id is not set for the tokenizer. Using eos_token_id as pad_token_id."
)
tokenizer.pad_token = tokenizer.eos_token
tokenizer.padding_side = "left"
gkw = {
"temperature": 0.5,
"top_p": 0.92,
"top_k": 0,
"max_new_tokens": 512,
"use_cache": True,
"do_sample": True,
"eos_token_id": tokenizer.eos_token_id,
"pad_token_id": tokenizer.pad_token_id,
"repetition_penalty": 1.1, # 1.0 means no penalty, > 1.0 means penalty, 1.2 from CTRL paper
}
def mpt_7b(s):
input_ids = tokenizer(s, return_tensors="pt").input_ids
input_ids = input_ids.to(model.device)
with torch.no_grad():
output_ids = model.generate(input_ids, **gkw)
# Slice the output_ids tensor to get only new tokens
new_tokens = output_ids[0, len(input_ids[0]) :]
output_text = tokenizer.decode(new_tokens, skip_special_tokens=True)
return output_text
|