Unverified Commit ba069a24 authored by Lianmin Zheng's avatar Lianmin Zheng Committed by GitHub
Browse files

Fix grammar backend (#2018)

parent 125b1199
...@@ -13,30 +13,5 @@ See the License for the specific language governing permissions and ...@@ -13,30 +13,5 @@ See the License for the specific language governing permissions and
limitations under the License. limitations under the License.
""" """
import json # TODO(lmzheng): make this an optional dependency
from typing import Dict, Optional, Union from sglang.srt.constrained.outlines_backend import build_regex_from_object
from pydantic import BaseModel
try:
from outlines.fsm.json_schema import build_regex_from_object
except ImportError:
# Since outlines 0.0.32, build_regex_from_object is replaced by build_regex_from_schema,
# which only accepts string schema as input.
from outlines.fsm.json_schema import build_regex_from_schema
def build_regex_from_object(
object: Union[str, BaseModel, Dict], whitespace_pattern: Optional[str] = None
):
if isinstance(object, type(BaseModel)):
schema = json.dumps(object.model_json_schema())
elif isinstance(object, Dict):
schema = json.dumps(object)
else:
schema = object
return build_regex_from_schema(schema, whitespace_pattern)
__all__ = [
"build_regex_from_object",
]
...@@ -95,9 +95,7 @@ class BaseToolCache: ...@@ -95,9 +95,7 @@ class BaseToolCache:
def get_cache_hit_rate(self): def get_cache_hit_rate(self):
with self.lock_metrics: with self.lock_metrics:
if self.metrics["total"] == 0: return self.metrics["hit"] / max(self.metrics["total"], 1)
return 0
return self.metrics["hit"] / self.metrics["total"]
def get_avg_init_time(self): def get_avg_init_time(self):
with self.lock_metrics: with self.lock_metrics:
......
"""
Copyright 2023-2024 SGLang Team
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
"""
"""Cache for the compressed finite state machine."""
import logging
from concurrent.futures import Future, ThreadPoolExecutor
from typing import List, Tuple, Union
import torch
from sglang.srt.constrained.outlines_cache import OutlinesCache, RegexGuide
from sglang.srt.constrained.outlines_jump_forward import (
OutlinesJumpCache,
OutlinesJumpForwardMap,
)
from sglang.srt.constrained.xgrammar_cache import (
GrammarMatcher,
XGrammarBackend,
XGrammarJumpCache,
)
logger = logging.getLogger(__name__)
class JumpHelper:
def __init__(
self, data: Union[List, str] = "", state: int = -1, suffix_ids=[]
) -> None:
self.data: Union[List, str] = data
self.state: int = state
self.suffix_ids: List[int] = suffix_ids
def can_jump(self):
return len(self.data) > 0
class Grammar:
def __init__(
self,
grammar: Union[GrammarMatcher, Tuple[RegexGuide, int]],
jump_map: Union[XGrammarJumpCache, OutlinesJumpForwardMap, None],
) -> None:
self.grammar = grammar
self.jump_map = jump_map
def accept_token(self, token: int):
if isinstance(self.grammar, GrammarMatcher):
assert self.grammar.accept_token(token)
else:
guide, state = self.grammar
self.grammar = guide, guide.get_next_state(state, token)
def try_jump(self, tokenizer) -> JumpHelper:
if isinstance(self.jump_map, XGrammarJumpCache):
assert isinstance(self.grammar, GrammarMatcher)
return JumpHelper(self.grammar.find_jump_forward_string())
elif isinstance(self.jump_map, OutlinesJumpForwardMap):
assert isinstance(self.grammar, Tuple)
_, state = self.grammar
jump_forward_bytes = self.jump_map.jump_forward_byte(state)
if jump_forward_bytes is None or len(jump_forward_bytes) == 0:
return JumpHelper() # can't jump
# preprocess the jump forward string
suffix_bytes = []
continuation_range = range(0x80, 0xC0)
cur_state = state
while (
len(jump_forward_bytes)
and jump_forward_bytes[0][0] in continuation_range
):
# continuation bytes
byte_edge = jump_forward_bytes.pop(0)
suffix_bytes.append(byte_edge[0])
cur_state = byte_edge[1]
suffix_tokens = [f"<0x{hex(b)[2:].upper()}>" for b in suffix_bytes]
suffix_ids = tokenizer.convert_tokens_to_ids(suffix_tokens)
return JumpHelper(suffix_ids, cur_state, suffix_bytes)
else:
return JumpHelper() # can't jump
def jump_forward_str_state(self, helper: JumpHelper) -> Tuple[str, int]:
if isinstance(helper.data, str):
return helper.data, -1
else:
assert isinstance(self.jump_map, OutlinesJumpForwardMap)
return self.jump_map.jump_forward_symbol(helper.state)
def jump_and_retokenize(
self, old_output_ids: List[int], new_output_ids: List[int], next_state: int
):
if isinstance(self.grammar, GrammarMatcher):
k = 0
for i, old_id in enumerate(old_output_ids):
if old_id == new_output_ids[i]:
k = i + 1
else:
break
# rollback to the last token that is the same
if k < len(old_output_ids):
self.grammar.rollback(len(old_output_ids) - k)
for i in range(k, len(new_output_ids)):
assert self.grammar.accept_token(new_output_ids[i])
else:
self.grammar = self.grammar[0], next_state
def fill_vocab_mask(self, vocab_mask: torch.Tensor, vocab_size: int):
if isinstance(self.grammar, GrammarMatcher):
# Note that this bitmask is a bitset, not bool
bitmask = self.grammar.get_next_token_bitmask()
# Mask the tokens that are not allowed
vocab_mask[
self.grammar.get_rejected_tokens_from_bitmask(bitmask, vocab_size)
] = 1
else:
guide, state = self.grammar
vocab_mask.fill_(1)
vocab_mask[guide.get_next_instruction(state).tokens] = 0
class GrammarBackend:
def __init__(
self,
tokenizer_path,
tokenizer_args_dict,
skip_tokenizer_init=False,
whitespace_patterns=None,
backend=None,
allow_jump=False,
):
self.executor = ThreadPoolExecutor()
self.backend = backend
if backend == "xgrammar":
self.grammar_cache = XGrammarBackend(
tokenizer_path=tokenizer_path,
tokenizer_args_dict=tokenizer_args_dict,
skip_tokenizer_init=skip_tokenizer_init,
whitespace_patterns=whitespace_patterns,
)
self.jump_cache = XGrammarJumpCache() if allow_jump else None
else:
assert backend == "outlines"
self.grammar_cache = OutlinesCache(
tokenizer_path=tokenizer_path,
tokenizer_args_dict=tokenizer_args_dict,
skip_tokenizer_init=skip_tokenizer_init,
constrained_json_whitespace_pattern=whitespace_patterns,
)
self.jump_cache = OutlinesJumpCache() if allow_jump else None
def _query(self, key: Tuple[str, str], vocab_size: int) -> Grammar:
if isinstance(self.grammar_cache, XGrammarBackend):
return Grammar(self.grammar_cache.query(key, vocab_size), self.jump_cache)
else:
guide, regex = self.grammar_cache.query(key)
jump_map = self.jump_cache.query(regex)
return Grammar((guide, 0), jump_map)
def query(self, key: Tuple[str, str], vocab_size: int) -> Future:
return self.executor.submit(self._query, key, vocab_size)
def reset(self):
self.grammar_cache.reset()
self.jump_cache.reset()
...@@ -13,41 +13,139 @@ See the License for the specific language governing permissions and ...@@ -13,41 +13,139 @@ See the License for the specific language governing permissions and
limitations under the License. limitations under the License.
""" """
"""Cache for the compressed finite state machine.""" """Constrained decoding with outlines backend."""
import json
import logging import logging
from concurrent.futures import Future, ThreadPoolExecutor
from typing import Dict, List, Optional, Tuple, Union
import torch
from interegular import InvalidSyntax, parse_pattern from interegular import InvalidSyntax, parse_pattern
from outlines.fsm.guide import RegexGuide from outlines.fsm.guide import RegexGuide
from outlines.models.transformers import TransformerTokenizer from outlines.models.transformers import TransformerTokenizer
from transformers import AutoTokenizer from pydantic import BaseModel
from sglang.srt.constrained import build_regex_from_object
from sglang.srt.constrained.base_tool_cache import BaseToolCache from sglang.srt.constrained.base_tool_cache import BaseToolCache
from sglang.srt.constrained.outlines_jump_forward import (
OutlinesJumpForwardCache,
OutlinesJumpForwardMap,
)
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
class OutlinesCache(BaseToolCache): try:
from outlines.fsm.json_schema import build_regex_from_object
except ImportError:
# Since outlines 0.0.32, build_regex_from_object is replaced by build_regex_from_schema,
# which only accepts string schema as input.
from outlines.fsm.json_schema import build_regex_from_schema
def build_regex_from_object(
object: Union[str, BaseModel, Dict], whitespace_pattern: Optional[str] = None
):
if isinstance(object, type(BaseModel)):
schema = json.dumps(object.model_json_schema())
elif isinstance(object, Dict):
schema = json.dumps(object)
else:
schema = object
return build_regex_from_schema(schema, whitespace_pattern)
class OutlinesGrammar:
def __init__( def __init__(
self, self,
tokenizer_path, guide: RegexGuide,
tokenizer_args_dict, state: int,
enable=True, jump_forward_map: Union[OutlinesJumpForwardMap, None],
skip_tokenizer_init=False, ) -> None:
constrained_json_whitespace_pattern=None, self.guide = guide
): self.state = state
super().__init__(enable=enable) self.jump_forward_map = jump_forward_map
if ( def accept_token(self, token: int):
skip_tokenizer_init self.state = self.guide.get_next_state(self.state, token)
or tokenizer_path.endswith(".json")
or tokenizer_path.endswith(".model") def try_jump_forward(self, tokenizer) -> Optional[Tuple]:
if not self.jump_forward_map:
return None
jump_forward_bytes = self.jump_forward_map.jump_forward_byte(self.state)
if jump_forward_bytes is None or len(jump_forward_bytes) <= 1:
return None
# preprocess the jump forward string
suffix_bytes = []
continuation_range = range(0x80, 0xC0)
cur_state = self.state
while (
len(jump_forward_bytes) and jump_forward_bytes[0][0] in continuation_range
): ):
# Do not support TiktokenTokenizer or SentencePieceTokenizer # continuation bytes
return byte_edge = jump_forward_bytes.pop(0)
suffix_bytes.append(byte_edge[0])
cur_state = byte_edge[1]
suffix_tokens = [f"<0x{hex(b)[2:].upper()}>" for b in suffix_bytes]
suffix_ids = tokenizer.convert_tokens_to_ids(suffix_tokens)
return suffix_ids, cur_state
def jump_forward_str_state(self, helper: Tuple[List[int], str]) -> Tuple[str, int]:
_, cur_state = helper
return self.jump_forward_map.jump_forward_symbol(cur_state)
def jump_and_retokenize(
self, old_output_ids: List[int], new_output_ids: List[int], next_state: int
):
self.state = next_state
def fill_vocab_mask(self, vocab_mask: torch.Tensor):
vocab_mask.fill_(1)
vocab_mask[self.guide.get_next_instruction(self.state).tokens] = 0
class OutlinesGrammarBackend:
def __init__(
self,
tokenizer,
whitespace_patterns: bool,
allow_jump_forward: bool,
):
self.executor = ThreadPoolExecutor()
self.grammar_cache = OutlinesCache(
tokenizer,
whitespace_pattern=whitespace_patterns,
)
self.jump_forward_cache = (
OutlinesJumpForwardCache() if allow_jump_forward else None
)
def _query(self, key: Tuple[str, str]) -> OutlinesGrammar:
guide, regex = self.grammar_cache.query(key)
jump_forward_map = (
self.jump_forward_cache.query(regex) if self.jump_forward_cache else None
)
return OutlinesGrammar(guide, 0, jump_forward_map)
def query(self, key: Tuple[str, str]) -> Future:
return self.executor.submit(self._query, key)
def reset(self):
self.grammar_cache.reset()
if self.jump_forward_cache:
self.jump_forward_cache.reset()
class OutlinesCache(BaseToolCache):
def __init__(
self,
tokenizer,
whitespace_pattern=None,
):
super().__init__(enable=True)
tokenizer_args_dict.setdefault("padding_side", "left")
tokenizer = AutoTokenizer.from_pretrained(tokenizer_path, **tokenizer_args_dict)
try: try:
self.outlines_tokenizer = TransformerTokenizer(tokenizer) self.outlines_tokenizer = TransformerTokenizer(tokenizer)
except AttributeError: except AttributeError:
...@@ -69,7 +167,7 @@ class OutlinesCache(BaseToolCache): ...@@ -69,7 +167,7 @@ class OutlinesCache(BaseToolCache):
self.outlines_tokenizer.vocabulary = ( self.outlines_tokenizer.vocabulary = (
self.outlines_tokenizer.tokenizer.get_vocab() self.outlines_tokenizer.tokenizer.get_vocab()
) )
self.constrained_json_whitespace_pattern = constrained_json_whitespace_pattern self.whitespace_pattern = whitespace_pattern
def init_value(self, key): def init_value(self, key):
key_type, key_string = key key_type, key_string = key
...@@ -77,7 +175,7 @@ class OutlinesCache(BaseToolCache): ...@@ -77,7 +175,7 @@ class OutlinesCache(BaseToolCache):
try: try:
regex = build_regex_from_object( regex = build_regex_from_object(
key_string, key_string,
whitespace_pattern=self.constrained_json_whitespace_pattern, whitespace_pattern=self.whitespace_pattern,
) )
except NotImplementedError as e: except NotImplementedError as e:
logger.warning( logger.warning(
...@@ -93,4 +191,13 @@ class OutlinesCache(BaseToolCache): ...@@ -93,4 +191,13 @@ class OutlinesCache(BaseToolCache):
except InvalidSyntax as e: except InvalidSyntax as e:
logger.warning(f"skip invalid regex guide: {regex=}, {e=}") logger.warning(f"skip invalid regex guide: {regex=}, {e=}")
return None, regex return None, regex
return RegexGuide(regex, self.outlines_tokenizer), regex
ret = RegexGuide(regex, self.outlines_tokenizer), regex
return ret
def _query(self, key: Tuple[str, str]):
guide, regex = self.grammar_cache.query(key)
jump_forward_map = (
self.jump_forward_cache.query(regex) if self.jump_forward_cache else None
)
return OutlinesGrammar(guide, 0, jump_forward_map)
...@@ -164,7 +164,7 @@ class OutlinesJumpForwardMap: ...@@ -164,7 +164,7 @@ class OutlinesJumpForwardMap:
) )
class OutlinesJumpCache(BaseToolCache): class OutlinesJumpForwardCache(BaseToolCache):
def __init__(self): def __init__(self):
super().__init__() super().__init__()
......
...@@ -3,7 +3,9 @@ Copyright 2023-2024 SGLang Team ...@@ -3,7 +3,9 @@ Copyright 2023-2024 SGLang Team
Licensed under the Apache License, Version 2.0 (the "License"); Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License. you may not use this file except in compliance with the License.
You may obtain a copy of the License at You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0 http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS, distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
...@@ -11,50 +13,98 @@ See the License for the specific language governing permissions and ...@@ -11,50 +13,98 @@ See the License for the specific language governing permissions and
limitations under the License. limitations under the License.
""" """
"""Cache for the compressed finite state machine.""" """Constrained decoding with xgrammar backend."""
from typing import Tuple from concurrent.futures import Future, ThreadPoolExecutor
from typing import List, Tuple
from transformers import AutoTokenizer import torch
try: try:
from xgrammar import CachedGrammarCompiler, CompiledGrammar, GrammarMatcher from xgrammar import CachedGrammarCompiler, CompiledGrammar, GrammarMatcher
import_error = None
except ImportError as e: except ImportError as e:
import_error = e
class Dummy: class Dummy:
pass pass
GrammarMatcher = Dummy GrammarMatcher = CompiledGrammar = CachedGrammarCompiler = Dummy
CompiledGrammar = Dummy
CachedGrammarCompiler = Dummy
MAX_ROLLBACK_TOKENS = 10 MAX_ROLLBACK_TOKENS = 10
class XGrammarJumpCache: class XGrammarGrammar:
"""A dummy class."""
def reset(self): def __init__(self, matcher: GrammarMatcher, vocab_size: int) -> None:
pass self.matcher = matcher
self.vocab_size = vocab_size
def accept_token(self, token: int):
assert self.matcher.accept_token(token)
def try_jump_forward(self, tokenizer) -> Tuple[List[int], str]:
return [], self.matcher.find_jump_forward_string()
class XGrammarBackend: def jump_forward_str_state(self, helper: Tuple[List[int], str]) -> Tuple[str, int]:
_, data = helper
return data, -1
def jump_and_retokenize(
self, old_output_ids: List[int], new_output_ids: List[int], next_state: int
):
k = 0
for i, old_id in enumerate(old_output_ids):
if old_id == new_output_ids[i]:
k = i + 1
else:
break
# rollback to the last token that is the same
if k < len(old_output_ids):
self.matcher.rollback(len(old_output_ids) - k)
for i in range(k, len(new_output_ids)):
assert self.matcher.accept_token(new_output_ids[i])
def fill_vocab_mask(self, vocab_mask: torch.Tensor):
# Note that this bitmask is a bitset, not bool
bitmask = self.matcher.get_next_token_bitmask()
# Mask the tokens that are not allowed
vocab_mask[
self.matcher.get_rejected_tokens_from_bitmask(bitmask, self.vocab_size)
] = 1
class XGrammarGrammarBackend:
def __init__( def __init__(
self, self,
tokenizer_path, tokenizer,
tokenizer_args_dict, vocab_size: int,
skip_tokenizer_init=False,
whitespace_patterns=None,
): ):
# TODO(dark): how to deal with whitespace_patterns and skip_tokenizer_init if import_error:
if skip_tokenizer_init: raise import_error
return
tokenizer = AutoTokenizer.from_pretrained(tokenizer_path, **tokenizer_args_dict) self.executor = ThreadPoolExecutor()
self.grammar_cache: CachedGrammarCompiler = CachedGrammarCompiler( self.grammar_cache = XGrammarCache(tokenizer, vocab_size)
tokenizer_or_vocab=tokenizer self.vocab_size = vocab_size
)
def _query(self, key: Tuple[str, str]) -> XGrammarGrammar:
return XGrammarGrammar(self.grammar_cache.query(key), self.vocab_size)
def query(self, key: Tuple[str, str]) -> Future:
return self.executor.submit(self._query, key)
def reset(self):
self.grammar_cache.reset()
class XGrammarCache:
def __init__(self, tokenizer, vocab_size: int):
self.grammar_cache = CachedGrammarCompiler(tokenizer_or_vocab=tokenizer)
self.vocab_size = vocab_size
def get_context(self, key: Tuple[str, str]) -> CompiledGrammar: def get_context(self, key: Tuple[str, str]) -> CompiledGrammar:
key_type, key_string = key key_type, key_string = key
...@@ -65,10 +115,12 @@ class XGrammarBackend: ...@@ -65,10 +115,12 @@ class XGrammarBackend:
else: else:
raise ValueError(f"Invalid key_type: {key_type}") raise ValueError(f"Invalid key_type: {key_type}")
def query(self, key: Tuple[str, str], vocab_size: int) -> GrammarMatcher: def query(self, key: Tuple[str, str]) -> GrammarMatcher:
ctx = self.get_context(key) ctx = self.get_context(key)
return GrammarMatcher( return GrammarMatcher(
ctx, max_rollback_tokens=MAX_ROLLBACK_TOKENS, mask_vocab_size=vocab_size ctx,
max_rollback_tokens=MAX_ROLLBACK_TOKENS,
mask_vocab_size=self.vocab_size,
) )
def reset(self): def reset(self):
......
...@@ -37,7 +37,6 @@ import torch ...@@ -37,7 +37,6 @@ import torch
from sglang.global_config import global_config from sglang.global_config import global_config
from sglang.srt.configs.model_config import ModelConfig from sglang.srt.configs.model_config import ModelConfig
from sglang.srt.constrained.grammar import Grammar
from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache
from sglang.srt.mem_cache.chunk_cache import ChunkCache from sglang.srt.mem_cache.chunk_cache import ChunkCache
from sglang.srt.mem_cache.memory_pool import BaseTokenToKVPool, ReqToTokenPool from sglang.srt.mem_cache.memory_pool import BaseTokenToKVPool, ReqToTokenPool
...@@ -249,7 +248,7 @@ class Req: ...@@ -249,7 +248,7 @@ class Req:
self.embedding = None self.embedding = None
# Constrained decoding # Constrained decoding
self.grammar: Optional[Grammar] = None self.grammar = None
# The number of cached tokens, that were already cached in the KV cache # The number of cached tokens, that were already cached in the KV cache
self.cached_tokens = 0 self.cached_tokens = 0
...@@ -359,8 +358,6 @@ class Req: ...@@ -359,8 +358,6 @@ class Req:
return return
def jump_forward_and_retokenize(self, jump_forward_str, next_state): def jump_forward_and_retokenize(self, jump_forward_str, next_state):
assert self.grammar is not None and self.tokenizer is not None
if self.origin_input_text is None: if self.origin_input_text is None:
# Recovering text can only use unpadded ids # Recovering text can only use unpadded ids
self.origin_input_text = self.tokenizer.decode( self.origin_input_text = self.tokenizer.decode(
...@@ -809,9 +806,10 @@ class ScheduleBatch: ...@@ -809,9 +806,10 @@ class ScheduleBatch:
for i, req in enumerate(self.reqs): for i, req in enumerate(self.reqs):
if req.grammar is not None: if req.grammar is not None:
jump_helper = req.grammar.try_jump(req.tokenizer) jump_helper = req.grammar.try_jump_forward(req.tokenizer)
if jump_helper.can_jump(): if jump_helper:
suffix_ids = jump_helper.suffix_ids suffix_ids, _ = jump_helper
# Current ids, for cache and revert # Current ids, for cache and revert
cur_all_ids = tuple(req.origin_input_ids + req.output_ids)[:-1] cur_all_ids = tuple(req.origin_input_ids + req.output_ids)[:-1]
cur_output_ids = req.output_ids cur_output_ids = req.output_ids
...@@ -827,6 +825,8 @@ class ScheduleBatch: ...@@ -827,6 +825,8 @@ class ScheduleBatch:
next_state, next_state,
) = req.grammar.jump_forward_str_state(jump_helper) ) = req.grammar.jump_forward_str_state(jump_helper)
# Make the incrementally decoded text part of jump_forward_str
# so that the UTF-8 will not corrupt
jump_forward_str = new_text + jump_forward_str jump_forward_str = new_text + jump_forward_str
if not req.jump_forward_and_retokenize( if not req.jump_forward_and_retokenize(
jump_forward_str, next_state jump_forward_str, next_state
......
...@@ -21,6 +21,7 @@ import threading ...@@ -21,6 +21,7 @@ import threading
import time import time
import warnings import warnings
from collections import deque from collections import deque
from concurrent import futures
from types import SimpleNamespace from types import SimpleNamespace
from typing import List, Optional from typing import List, Optional
...@@ -29,7 +30,6 @@ import zmq ...@@ -29,7 +30,6 @@ import zmq
from sglang.global_config import global_config from sglang.global_config import global_config
from sglang.srt.configs.model_config import ModelConfig from sglang.srt.configs.model_config import ModelConfig
from sglang.srt.constrained.grammar import GrammarBackend
from sglang.srt.hf_transformers_utils import get_processor, get_tokenizer from sglang.srt.hf_transformers_utils import get_processor, get_tokenizer
from sglang.srt.layers.logits_processor import LogitsProcessorOutput from sglang.srt.layers.logits_processor import LogitsProcessorOutput
from sglang.srt.managers.io_struct import ( from sglang.srt.managers.io_struct import (
...@@ -100,7 +100,7 @@ class Scheduler: ...@@ -100,7 +100,7 @@ class Scheduler:
self.tp_rank = tp_rank self.tp_rank = tp_rank
self.tp_size = server_args.tp_size self.tp_size = server_args.tp_size
self.schedule_policy = server_args.schedule_policy self.schedule_policy = server_args.schedule_policy
self.disable_regex_jump_forward = server_args.disable_regex_jump_forward self.disable_jump_forward = server_args.disable_jump_forward
self.lora_paths = server_args.lora_paths self.lora_paths = server_args.lora_paths
self.max_loras_per_batch = server_args.max_loras_per_batch self.max_loras_per_batch = server_args.max_loras_per_batch
self.enable_overlap = server_args.enable_overlap_schedule self.enable_overlap = server_args.enable_overlap_schedule
...@@ -234,22 +234,33 @@ class Scheduler: ...@@ -234,22 +234,33 @@ class Scheduler:
self.chunked_prefill_size is not None and server_args.enable_mixed_chunk self.chunked_prefill_size is not None and server_args.enable_mixed_chunk
) )
# Init the grammar cache for constrained generation # Init the grammar backend for constrained generation
self.grammar_cache = None
self.grammar_queue: List[Req] = [] self.grammar_queue: List[Req] = []
if not server_args.skip_tokenizer_init: if not server_args.skip_tokenizer_init:
self.grammar_cache = GrammarBackend( if server_args.grammar_backend == "outlines":
server_args.tokenizer_path, from sglang.srt.constrained.outlines_backend import (
{ OutlinesGrammarBackend,
"tokenizer_mode": server_args.tokenizer_mode, )
"trust_remote_code": server_args.trust_remote_code,
}, self.grammar_backend = OutlinesGrammarBackend(
skip_tokenizer_init=server_args.skip_tokenizer_init, self.tokenizer,
whitespace_patterns=server_args.constrained_json_whitespace_pattern, whitespace_patterns=server_args.constrained_json_whitespace_pattern,
backend=server_args.grammar_backend, allow_jump_forward=not server_args.disable_jump_forward,
allow_jump=not server_args.disable_regex_jump_forward, )
) elif server_args.grammar_backend == "xgrammar":
from sglang.srt.constrained.xgrammar_backend import (
XGrammarGrammarBackend,
)
self.grammar_backend = XGrammarGrammarBackend(
self.tokenizer, vocab_size=self.model_config.vocab_size
)
else:
raise ValueError(
f"Invalid grammar backend: {server_args.grammar_backend}"
)
else:
self.grammar_backend = None
# Init new token estimation # Init new token estimation
assert ( assert (
...@@ -461,15 +472,14 @@ class Scheduler: ...@@ -461,15 +472,14 @@ class Scheduler:
req.sampling_params.json_schema is not None req.sampling_params.json_schema is not None
or req.sampling_params.regex is not None or req.sampling_params.regex is not None
): ):
assert self.grammar_cache is not None assert self.grammar_backend is not None
if req.sampling_params.json_schema is not None: if req.sampling_params.json_schema is not None:
req.grammar = self.grammar_cache.query( req.grammar = self.grammar_backend.query(
("json", req.sampling_params.json_schema), ("json", req.sampling_params.json_schema),
self.model_config.vocab_size,
) )
elif req.sampling_params.regex is not None: elif req.sampling_params.regex is not None:
req.grammar = self.grammar_cache.query( req.grammar = self.grammar_backend.query(
("regex", req.sampling_params.regex), self.model_config.vocab_size ("regex", req.sampling_params.regex)
) )
# Truncate prompts that are too long # Truncate prompts that are too long
...@@ -638,14 +648,14 @@ class Scheduler: ...@@ -638,14 +648,14 @@ class Scheduler:
return self.running_batch return self.running_batch
def get_new_batch_prefill(self) -> Optional[ScheduleBatch]: def get_new_batch_prefill(self) -> Optional[ScheduleBatch]:
# Check if the grammar queue is ready # Check if the grammar is ready in the grammar queue
if self.grammar_queue: if self.grammar_queue:
new_grammar_queue = [] new_grammar_queue = []
for req in self.grammar_queue: for req in self.grammar_queue:
if req.grammar.done(): try:
req.grammar = req.grammar.result() req.grammar = req.grammar.result(timeout=0.05)
self.waiting_queue.append(req) self.waiting_queue.append(req)
else: except futures._base.TimeoutError:
new_grammar_queue.append(req) new_grammar_queue.append(req)
self.grammar_queue = new_grammar_queue self.grammar_queue = new_grammar_queue
...@@ -783,7 +793,7 @@ class Scheduler: ...@@ -783,7 +793,7 @@ class Scheduler:
) )
# Check for jump-forward # Check for jump-forward
if not self.disable_regex_jump_forward: if not self.disable_jump_forward:
jump_forward_reqs = batch.check_for_jump_forward(self.pad_input_ids_func) jump_forward_reqs = batch.check_for_jump_forward(self.pad_input_ids_func)
self.waiting_queue.extend(jump_forward_reqs) self.waiting_queue.extend(jump_forward_reqs)
if batch.is_empty(): if batch.is_empty():
...@@ -1142,8 +1152,8 @@ class Scheduler: ...@@ -1142,8 +1152,8 @@ class Scheduler:
): ):
self.tree_cache.reset() self.tree_cache.reset()
self.tree_cache_metrics = {"total": 0, "hit": 0} self.tree_cache_metrics = {"total": 0, "hit": 0}
if self.grammar_cache is not None: if self.grammar_backend is not None:
self.grammar_cache.reset() self.grammar_backend.reset()
# TODO(dark): reset the bnf cache # TODO(dark): reset the bnf cache
self.req_to_token_pool.clear() self.req_to_token_pool.clear()
self.token_to_kv_pool.clear() self.token_to_kv_pool.clear()
......
...@@ -6,7 +6,6 @@ from typing import TYPE_CHECKING, List, Optional ...@@ -6,7 +6,6 @@ from typing import TYPE_CHECKING, List, Optional
import torch import torch
import sglang.srt.sampling.penaltylib as penaltylib import sglang.srt.sampling.penaltylib as penaltylib
from sglang.srt.constrained.grammar import Grammar
if TYPE_CHECKING: if TYPE_CHECKING:
from sglang.srt.managers.schedule_batch import ScheduleBatch from sglang.srt.managers.schedule_batch import ScheduleBatch
...@@ -31,7 +30,7 @@ class SamplingBatchInfo: ...@@ -31,7 +30,7 @@ class SamplingBatchInfo:
logit_bias: torch.Tensor = None logit_bias: torch.Tensor = None
vocab_mask: Optional[torch.Tensor] = None vocab_mask: Optional[torch.Tensor] = None
grammars: Optional[List[Optional[Grammar]]] = None grammars: Optional[List] = None
# Penalizer # Penalizer
penalizer_orchestrator: Optional[penaltylib.BatchedPenalizerOrchestrator] = None penalizer_orchestrator: Optional[penaltylib.BatchedPenalizerOrchestrator] = None
...@@ -146,7 +145,7 @@ class SamplingBatchInfo: ...@@ -146,7 +145,7 @@ class SamplingBatchInfo:
) )
for i, grammar in enumerate(self.grammars): for i, grammar in enumerate(self.grammars):
if grammar is not None: if grammar is not None:
grammar.fill_vocab_mask(self.vocab_mask[i], self.vocab_size) grammar.fill_vocab_mask(self.vocab_mask[i])
def filter_batch(self, unfinished_indices: List[int], new_indices: torch.Tensor): def filter_batch(self, unfinished_indices: List[int], new_indices: torch.Tensor):
if self.penalizer_orchestrator: if self.penalizer_orchestrator:
......
...@@ -111,7 +111,7 @@ class ServerArgs: ...@@ -111,7 +111,7 @@ class ServerArgs:
disable_flashinfer: bool = False disable_flashinfer: bool = False
disable_flashinfer_sampling: bool = False disable_flashinfer_sampling: bool = False
disable_radix_cache: bool = False disable_radix_cache: bool = False
disable_regex_jump_forward: bool = False disable_jump_forward: bool = False
disable_cuda_graph: bool = False disable_cuda_graph: bool = False
disable_cuda_graph_padding: bool = False disable_cuda_graph_padding: bool = False
disable_disk_cache: bool = False disable_disk_cache: bool = False
...@@ -574,7 +574,7 @@ class ServerArgs: ...@@ -574,7 +574,7 @@ class ServerArgs:
type=str, type=str,
choices=["xgrammar", "outlines"], choices=["xgrammar", "outlines"],
default=ServerArgs.grammar_backend, default=ServerArgs.grammar_backend,
help="Choose the backend for constrained decoding.", help="Choose the backend for grammar-guided decoding.",
) )
# Optimization/debug options # Optimization/debug options
...@@ -594,9 +594,9 @@ class ServerArgs: ...@@ -594,9 +594,9 @@ class ServerArgs:
help="Disable RadixAttention for prefix caching.", help="Disable RadixAttention for prefix caching.",
) )
parser.add_argument( parser.add_argument(
"--disable-regex-jump-forward", "--disable-jump-forward",
action="store_true", action="store_true",
help="Disable regex jump-forward.", help="Disable jump-forward for grammar-guided decoding.",
) )
parser.add_argument( parser.add_argument(
"--disable-cuda-graph", "--disable-cuda-graph",
...@@ -616,7 +616,6 @@ class ServerArgs: ...@@ -616,7 +616,6 @@ class ServerArgs:
parser.add_argument( parser.add_argument(
"--disable-custom-all-reduce", "--disable-custom-all-reduce",
action="store_true", action="store_true",
default=False,
help="Disable the custom all-reduce kernel and fall back to NCCL.", help="Disable the custom all-reduce kernel and fall back to NCCL.",
) )
parser.add_argument( parser.add_argument(
...@@ -688,7 +687,6 @@ class ServerArgs: ...@@ -688,7 +687,6 @@ class ServerArgs:
) )
parser.add_argument( parser.add_argument(
"--delete-ckpt-after-loading", "--delete-ckpt-after-loading",
default=ServerArgs.delete_ckpt_after_loading,
action="store_true", action="store_true",
help="Delete the model checkpoint after loading the model.", help="Delete the model checkpoint after loading the model.",
) )
......
...@@ -61,18 +61,27 @@ class TestJSONConstrained(unittest.TestCase): ...@@ -61,18 +61,27 @@ class TestJSONConstrained(unittest.TestCase):
"logprob_start_len": 0, "logprob_start_len": 0,
}, },
) )
print(json.dumps(response.json())) ret = response.json()
print(json.dumps(ret))
print("=" * 100) print("=" * 100)
if not json_schema: if not json_schema:
return return
# Make sure the json output is valid
try: try:
js_obj = json.loads(response.json()["text"]) js_obj = json.loads(ret["text"])
except (TypeError, json.decoder.JSONDecodeError): except (TypeError, json.decoder.JSONDecodeError):
raise raise
assert isinstance(js_obj["name"], str)
assert isinstance(js_obj["population"], int) self.assertIsInstance(js_obj["name"], str)
self.assertIsInstance(js_obj["population"], int)
# Make sure jump forward is triggered
self.assertGreater(
ret["meta_info"]["completion_tokens"],
ret["meta_info"]["completion_tokens_wo_jump_forward"],
)
def test_json_generate(self): def test_json_generate(self):
self.run_decode(json_schema=self.json_schema) self.run_decode(json_schema=self.json_schema)
...@@ -100,8 +109,9 @@ class TestJSONConstrained(unittest.TestCase): ...@@ -100,8 +109,9 @@ class TestJSONConstrained(unittest.TestCase):
except (TypeError, json.decoder.JSONDecodeError): except (TypeError, json.decoder.JSONDecodeError):
print("JSONDecodeError", text) print("JSONDecodeError", text)
raise raise
assert isinstance(js_obj["name"], str), f"{js_obj=}"
assert isinstance(js_obj["population"], int), f"{js_obj=}" self.assertIsInstance(js_obj["name"], str)
self.assertIsInstance(js_obj["population"], int)
def test_mix_json_and_other(self): def test_mix_json_and_other(self):
json_schemas = [None, None, self.json_schema, self.json_schema] * 10 json_schemas = [None, None, self.json_schema, self.json_schema] * 10
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment