MCPcopy Index your code
hub / github.com/PaddlePaddle/FastDeploy / FDConfig

Class FDConfig

fastdeploy/config.py:1698–2267  ·  view source on GitHub ↗

The configuration class which contains all fastdeploy-related configuration. This simplifies passing around the distinct configurations in the codebase.

Source from the content-addressed store, hash-verified

1696
1697
1698class FDConfig:
1699 """
1700 The configuration class which contains all fastdeploy-related configuration. This
1701 simplifies passing around the distinct configurations in the codebase.
1702 """
1703
1704 def __init__(
1705 self,
1706 model_config: ModelConfig = None,
1707 cache_config: CacheConfig = None,
1708 parallel_config: ParallelConfig = None,
1709 load_config: LoadConfig = None,
1710 commit_config: CommitConfig = CommitConfig(),
1711 scheduler_config: SchedulerConfig = None,
1712 device_config: DeviceConfig = None,
1713 quant_config: QuantConfigBase = None,
1714 graph_opt_config: GraphOptimizationConfig = None,
1715 plas_attention_config: PlasAttentionConfig = None,
1716 speculative_config: SpeculativeConfig = None,
1717 eplb_config: EPLBConfig = None,
1718 structured_outputs_config: StructuredOutputsConfig = None,
1719 router_config: RouterConfig = None,
1720 tokenizer: str = None,
1721 ips: str = None,
1722 use_warmup: bool = False,
1723 limit_mm_per_prompt: Optional[Dict[str, Any]] = None,
1724 mm_processor_kwargs: Optional[Dict[str, Any]] = None,
1725 max_num_partial_prefills: int = 1,
1726 max_long_partial_prefills: int = 1,
1727 long_prefill_token_threshold: int = 0,
1728 early_stop_config: Optional[Dict[str, Any]] = None,
1729 tool_parser: str = None,
1730 test_mode=False,
1731 routing_replay_config: Optional[RoutingReplayConfig] = None,
1732 deploy_modality: "DeployModality" = None,
1733 ):
1734 self.model_config: ModelConfig = model_config # type: ignore
1735 self.cache_config: CacheConfig = cache_config # type: ignore
1736 self.scheduler_config: SchedulerConfig = scheduler_config # type: ignore
1737 self.parallel_config = parallel_config # type: ignore
1738 self.speculative_config: SpeculativeConfig = speculative_config
1739 self.eplb_config: Optional[EPLBConfig] = eplb_config
1740 self.device_config: DeviceConfig = device_config # type: ignore
1741 self.load_config: LoadConfig = load_config
1742 self.quant_config: Optional[QuantConfigBase] = quant_config
1743 self.graph_opt_config: Optional[GraphOptimizationConfig] = graph_opt_config
1744 self.early_stop_config: Optional[EarlyStopConfig] = early_stop_config
1745 self.plas_attention_config: Optional[PlasAttentionConfig] = plas_attention_config
1746 self.structured_outputs_config: StructuredOutputsConfig = structured_outputs_config
1747 self.router_config: RouterConfig = router_config
1748 self.routing_replay_config = routing_replay_config
1749 self.deploy_modality: DeployModality = deploy_modality if deploy_modality is not None else DeployModality.MIXED
1750 max_capture_shape = self.scheduler_config.max_num_seqs
1751 if self.speculative_config is not None and self.speculative_config.method in ["mtp", "suffix"]:
1752 max_capture_shape = self.scheduler_config.max_num_seqs * (
1753 self.speculative_config.num_speculative_tokens + 1
1754 )
1755 assert max_capture_shape % 2 == 0, "CUDAGraph only supports capturing even token nums in MTP scenarios."

Callers 15

__init__Method · 0.90
create_engine_configMethod · 0.90
initialize_fd_configFunction · 0.90
test_fdconfig_nnodeMethod · 0.90
test_fdconfig_ipsMethod · 0.90
real_fd_configFunction · 0.90
mock_fd_configFunction · 0.90
mock_fd_config_tp2Function · 0.90

Calls

no outgoing calls

Tested by 15

test_fdconfig_nnodeMethod · 0.72
test_fdconfig_ipsMethod · 0.72
real_fd_configFunction · 0.72
mock_fd_configFunction · 0.72
mock_fd_config_tp2Function · 0.72
mock_fd_config_qwen3Function · 0.72
get_fd_configFunction · 0.72
__init__Method · 0.72