Refactor DSpark disagg

[zhangxiaolei123456]

Motivation

Modifications

Accuracy Tests

Speed Tests and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
• Follow the SGLang code style guidance.

Review and Merge Process

1. Ping Merge Oncalls to start the process. See the PR Merge Process.
2. Get approvals from CODEOWNERS and other reviewers.
3. Trigger CI tests with comments or contact authorized users to do so.
   • Common commands include /tag-and-rerun-ci, /tag-run-ci-label, /rerun-failed-ci
4. After green CI and required approvals, ask Merge Oncalls or people with Write permission to merge the PR.

---

CI States

Latest PR Test (Base): ❌ Run #29752805296
Latest PR Test (Extra): ❌ Run #29752805120

[gemini-code-assist]

Code Review

This pull request refactors the disaggregation metadata configuration and hidden bootstrap plan resolution by extracting the logic into a new module, 'dspark_disaggregation.py', and renaming several 'DSpark' prefixed classes to 'PD' equivalents. The review feedback suggests improving robustness in the new module by converting string enum values to lowercase before comparison and using 'or []' or 'or 0' instead of '.get(..., [])' or '.get(..., 0)' to safely handle cases where dictionary values might be explicitly set to 'None'.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

To ensure robustness against case-sensitivity issues, it is recommended to convert mode_value to lowercase before checking if it is in the allowed set of modes. This prevents potential runtime issues if the enum values or strings are passed in uppercase or mixed-case.

    mode_value = getattr(disaggregation_mode, "value", disaggregation_mode)
    if isinstance(mode_value, str):
        mode_value = mode_value.lower()
    if mode_value not in ("decode", "prefill"):
        return DSparkDisaggMetadataConfig()

[gemini-code-assist]

Similarly, backend_value should be converted to lowercase to handle potential case mismatches robustly.

Suggested change

	backend_value = getattr(transfer_backend, "value", transfer_backend)
	if backend_value not in ("mooncake", "fake"):
	raise NotImplementedError(
	"DSpark PD hidden transfer is implemented only for Mooncake/Fake "
	f"backends, got {backend_value}."
	)
	backend_value = getattr(transfer_backend, "value", transfer_backend)
	if isinstance(backend_value, str):
	backend_value = backend_value.lower()
	if backend_value not in ("mooncake", "fake"):
	raise NotImplementedError(
	"DSpark PD hidden transfer is implemented only for Mooncake/Fake "
	f"backends, got {backend_value}."
	)

[gemini-code-assist]

Using .get(..., []) or .get(..., 0) can return None if the key exists in the dictionary but its value is explicitly set to None (e.g., from a parsed JSON null). To prevent a TypeError during iteration or type conversion, use or [] and or 0 instead.

Suggested change

	local_layer_ids = (
	[int(x) for x in local_pp_slice.get("layer_ids", [])]
	if local_pp_slice
	else (
	[]
	if pp_slices
	else [int(x) for x in metadata.get("target_layer_ids", [])]
	)
	)
	local_slice_len = (
	int(local_pp_slice.get("slice_len", 0))
	if local_pp_slice
	else len(local_layer_ids) * int(model_config.hidden_size)
	)
	local_layer_ids = (
	[int(x) for x in (local_pp_slice.get("layer_ids") or [])]
	if local_pp_slice
	else (
	[]
	if pp_slices
	else [int(x) for x in (metadata.get("target_layer_ids") or [])]
	)
	)
	local_slice_len = (
	int(local_pp_slice.get("slice_len") or 0)
	if local_pp_slice
	else len(local_layer_ids) * int(model_config.hidden_size)
	)

[gemini-code-assist]

Similarly, use or [] instead of .get(..., []) to safely handle cases where dst_indices might be explicitly set to None in the metadata.

Suggested change

	dst_indices = [
	int(x)
	for x in (
	local_pp_slice.get("dst_indices", [])
	if local_pp_slice
	else metadata.get("dst_indices", [])
	)
	]
	dst_indices = [
	int(x)
	for x in (
	local_pp_slice.get("dst_indices") or []
	if local_pp_slice
	else metadata.get("dst_indices") or []
	)
	]