[FDConfig]Remove max_num_batched_tokens/max_num_seqs in parallel config

[YuanRisheng]

移除parallel_config和FDConfig中的max_num_batched_tokens和max_num_seqs，将其放入SchedulerConfig

[paddle-bot]

Thanks for your contribution!

[Copilot]

Pull Request Overview

This PR moves max_num_batched_tokens and max_num_seqs configuration parameters from ParallelConfig and FDConfig to a dedicated SchedulerConfig class. This refactoring improves the separation of concerns by placing scheduler-related configuration parameters in their appropriate config class.

• Introduces SchedulerConfig import and instantiation across test files and worker components
• Updates all references to use scheduler_config.max_num_batched_tokens and scheduler_config.max_num_seqs instead of parallel_config equivalents
• Modifies config initialization in args_utils to properly pass parameters to SchedulerConfig

Reviewed Changes

Copilot reviewed 30 out of 30 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tests/v1/test_schedule_output.py	Updates test to use SchedulerConfig and new config structure
tests/v1/test_prefix_cache.py	Updates test configuration to use SchedulerConfig
tests/utils/test_config.py	Updates test assertions to access scheduler config properties
tests/utils.py	Moves max_num_seqs from parallel_config to scheduler_config
tests/graph_optimization/*.py	Updates multiple graph optimization tests to use SchedulerConfig
fastdeploy/worker/*.py	Updates worker components to access scheduler config properties
fastdeploy/spec_decode/base.py	Updates to use scheduler_config for max_num_seqs
fastdeploy/scheduler/config.py	Adds max_num_batched_tokens and max_num_seqs to SchedulerConfig
fastdeploy/model_executor/models/deepseek_v3.py	Updates buffer allocation to use scheduler config
fastdeploy/model_executor/layers/sample/sampler.py	Updates early stopper initialization
fastdeploy/model_executor/layers/backends/gcu/attention/*.py	Updates attention backends to use scheduler config
fastdeploy/model_executor/layers/attention/*.py	Updates attention components to use scheduler config
fastdeploy/model_executor/guided_decoding/xgrammar_backend.py	Updates batch size configuration
fastdeploy/engine/*.py	Updates engine components to use scheduler config
fastdeploy/config.py	Removes max_num_batched_tokens and max_num_seqs from ParallelConfig and FDConfig, adds SchedulerConfig integration

[Copilot]

This should use self.scheduler_config.max_num_seqs instead of self.max_num_seqs for consistency and to ensure the scheduler config values are used throughout.

Suggested change

	self.cache_config.postprocess(self.scheduler_config.max_num_batched_tokens, self.max_num_seqs)
	self.cache_config.postprocess(self.scheduler_config.max_num_batched_tokens, self.scheduler_config.max_num_seqs)

[YuanRisheng]

done