vllm.models.deepseek_v4.cpu.dspark

CPU DeepSeek-V4 DSpark (speculative decoding) — deferred, not implemented.

DSparkDeepseekV4ForCausalLM is only instantiated when the speculative- decoding registry resolves it by name for a DSpark speculative config; a plain (non-spec) DeepSeek-V4 CPU model never constructs this class. This stub exists solely so vllm.models.deepseek_v4.cpu is import-clean.