Spaces:
Runtime error
Runtime error
File size: 2,083 Bytes
c7a96cd |
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 57 58 59 60 61 62 63 64 65 66 67 68 69 |
import torch
import torch.distributed
from typing import Optional, List
from transformers import AutoTokenizer, AutoModelForCausalLM
from text_generation_server.models import CausalLM
FIM_PREFIX = "<fim-prefix>"
FIM_MIDDLE = "<fim-middle>"
FIM_SUFFIX = "<fim-suffix>"
FIM_PAD = "<fim-pad>"
EOD = "<|endoftext|>"
class SantaCoder(CausalLM):
def __init__(self, model_id: str, revision: Optional[str] = None, quantize=False):
if torch.cuda.is_available():
device = torch.device("cuda")
dtype = torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float32
else:
if quantize:
raise ValueError("quantization is not available on CPU")
device = torch.device("cpu")
dtype = torch.float32
tokenizer = AutoTokenizer.from_pretrained(
model_id, revision=revision, padding_side="left", truncation_side="left"
)
tokenizer.add_special_tokens(
{
"additional_special_tokens": [
EOD,
FIM_PREFIX,
FIM_MIDDLE,
FIM_SUFFIX,
FIM_PAD,
],
"pad_token": EOD,
}
)
self.model = (
AutoModelForCausalLM.from_pretrained(
model_id,
revision=revision,
torch_dtype=dtype,
load_in_8bit=quantize,
trust_remote_code=True, # required
)
.to(device)
.eval()
)
super(CausalLM, self).__init__(
tokenizer=tokenizer,
requires_padding=True,
dtype=dtype,
device=device,
decode_buffer=1,
)
def decode(self, generated_ids: List[int]) -> str:
# Do not skip special tokens as they are used for custom parsing rules of the generated text
return self.tokenizer.decode(
generated_ids, skip_special_tokens=False, cleanup_tokenization_spaces=False
)
|