Russell Bryant
e489ad7a21
- **Add SPDX license headers to python source files** - **Check for SPDX headers using pre-commit** commit 9d7ef44c3cfb72ca4c32e1c677d99259d10d4745 Author: Russell Bryant <rbryant@redhat.com> Date: Fri Jan 31 14:18:24 2025 -0500 Add SPDX license headers to python source files This commit adds SPDX license headers to python source files as recommended to the project by the Linux Foundation. These headers provide a concise way that is both human and machine readable for communicating license information for each source file. It helps avoid any ambiguity about the license of the code and can also be easily used by tools to help manage license compliance. The Linux Foundation runs license scans against the codebase to help ensure we are in compliance with the licenses of the code we use, including dependencies. Having these headers in place helps that tool do its job. More information can be found on the SPDX site: - https://spdx.dev/learn/handling-license-info/ Signed-off-by: Russell Bryant <rbryant@redhat.com> commit 5a1cf1cb3b80759131c73f6a9dddebccac039dea Author: Russell Bryant <rbryant@redhat.com> Date: Fri Jan 31 14:36:32 2025 -0500 Check for SPDX headers using pre-commit Signed-off-by: Russell Bryant <rbryant@redhat.com> --------- Signed-off-by: Russell Bryant <rbryant@redhat.com>
109 lines
4.5 KiB
Python
109 lines
4.5 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
|
|
from typing import List, Optional
|
|
|
|
from vllm.config import TokenizerPoolConfig
|
|
from vllm.lora.request import LoRARequest
|
|
from vllm.transformers_utils.tokenizer import (AnyTokenizer, encode_tokens,
|
|
get_lora_tokenizer,
|
|
get_lora_tokenizer_async,
|
|
get_tokenizer)
|
|
from vllm.utils import LRUCache
|
|
|
|
from .base_tokenizer_group import BaseTokenizerGroup
|
|
|
|
|
|
class TokenizerGroup(BaseTokenizerGroup):
|
|
"""A group of tokenizers that can be used for LoRA adapters."""
|
|
|
|
def __init__(self, tokenizer_id: str, enable_lora: bool, max_num_seqs: int,
|
|
max_input_length: Optional[int], **tokenizer_config):
|
|
self.tokenizer_id = tokenizer_id
|
|
self.tokenizer_config = tokenizer_config
|
|
self.enable_lora = enable_lora
|
|
self.max_input_length = max_input_length
|
|
self.tokenizer = get_tokenizer(self.tokenizer_id, **tokenizer_config)
|
|
max_loras = tokenizer_config.get("max_loras", 0)
|
|
self.lora_tokenizers = LRUCache[int, AnyTokenizer](
|
|
capacity=max(max_loras, max_num_seqs) if enable_lora else 0)
|
|
|
|
@classmethod
|
|
def from_config(cls, tokenizer_pool_config: Optional[TokenizerPoolConfig],
|
|
**init_kwargs) -> "TokenizerGroup":
|
|
return cls(**init_kwargs)
|
|
|
|
def ping(self) -> bool:
|
|
"""Check if the tokenizer group is alive."""
|
|
return True
|
|
|
|
def get_max_input_len(self,
|
|
lora_request: Optional[LoRARequest] = None
|
|
) -> Optional[int]:
|
|
"""Get the maximum input length for the LoRA request."""
|
|
return self.max_input_length
|
|
|
|
def _raise_if_input_too_long(self,
|
|
encoded_tokens: List[int],
|
|
lora_request: Optional[LoRARequest] = None):
|
|
input_length = len(encoded_tokens)
|
|
if lora_request:
|
|
max_input_length = (lora_request.long_lora_max_len
|
|
or self.max_input_length)
|
|
else:
|
|
max_input_length = self.max_input_length
|
|
if max_input_length is not None and input_length > max_input_length:
|
|
raise ValueError("Input too long.", input_length, max_input_length)
|
|
|
|
def encode(self,
|
|
prompt: str,
|
|
request_id: Optional[str] = None,
|
|
lora_request: Optional[LoRARequest] = None,
|
|
add_special_tokens: Optional[bool] = None) -> List[int]:
|
|
tokenizer = self.get_lora_tokenizer(lora_request)
|
|
ret = encode_tokens(tokenizer,
|
|
prompt,
|
|
add_special_tokens=add_special_tokens)
|
|
self._raise_if_input_too_long(ret, lora_request)
|
|
return ret
|
|
|
|
async def encode_async(
|
|
self,
|
|
prompt: str,
|
|
request_id: Optional[str] = None,
|
|
lora_request: Optional[LoRARequest] = None,
|
|
add_special_tokens: Optional[bool] = None) -> List[int]:
|
|
tokenizer = await self.get_lora_tokenizer_async(lora_request)
|
|
ret = encode_tokens(tokenizer,
|
|
prompt,
|
|
add_special_tokens=add_special_tokens)
|
|
self._raise_if_input_too_long(ret, lora_request)
|
|
return ret
|
|
|
|
def get_lora_tokenizer(
|
|
self,
|
|
lora_request: Optional[LoRARequest] = None,
|
|
) -> AnyTokenizer:
|
|
if not lora_request or not self.enable_lora:
|
|
return self.tokenizer
|
|
if lora_request.lora_int_id not in self.lora_tokenizers:
|
|
tokenizer = (get_lora_tokenizer(
|
|
lora_request, **self.tokenizer_config) or self.tokenizer)
|
|
self.lora_tokenizers.put(lora_request.lora_int_id, tokenizer)
|
|
return tokenizer
|
|
else:
|
|
return self.lora_tokenizers[lora_request.lora_int_id]
|
|
|
|
async def get_lora_tokenizer_async(
|
|
self,
|
|
lora_request: Optional[LoRARequest] = None,
|
|
) -> AnyTokenizer:
|
|
if not lora_request or not self.enable_lora:
|
|
return self.tokenizer
|
|
if lora_request.lora_int_id not in self.lora_tokenizers:
|
|
tokenizer = (await get_lora_tokenizer_async(
|
|
lora_request, **self.tokenizer_config) or self.tokenizer)
|
|
self.lora_tokenizers.put(lora_request.lora_int_id, tokenizer)
|
|
return tokenizer
|
|
else:
|
|
return self.lora_tokenizers[lora_request.lora_int_id]
|