Gaudi HPU Graphs: Static Shape Capture and Replay

HPU graphs are Gaudi's equivalent of CUDA graphs: they capture a sequence of kernel launches into a single graph that can be replayed with minimal launch overhead. This is critical for LLM inference where many small kernel launches (attention, FFN, sampling) would otherwise be launch-bound. The vLLM Gaudi plugin implements custom graph caching for fast warm-up and efficient memory use. torch.compile with mode='reduce-overhead' enables HPU graphs automatically. The key constraint is static shapes: the batch size and sequence length must be fixed at capture time, so any change triggers recompilation. This is why vLLM on Gaudi uses a bucketing mechanism to group requests into fixed-size buckets. The HabanaAI/vllm-fork reached EOL at v1.23.0; the vllm-gaudi plugin is the production path. HPU graphs are not portable to CUDA or ROCm; they are a Gaudi-specific feature that requires the Habana runtime (torch-hpu). RESOLVED (2026-10-04 cross-reference repair): affected_by_hardware values 'gaudi-2' -> intel-gaudi2 and 'gaudi-3' -> intel-gaudi3. Each value is the architecture string of exactly one accelerator record and this repo holds one card per Gaudi generation, so the mapping is forced - note it is forced by the repo's contents, not by the source document. RESOLVED 2026-10-04 by DELETION, not by repointing, and the deletion is the correct action rather than a dodge. 'hpu' names the Habana Processing Unit - the DEVICE CLASS spanning every Gaudi generation - not any card. This field already carried BOTH real targets, intel-gaudi2 and intel-gaudi3, before this pass, so 'hpu' was a third entry adding no hardware scope that was not already claimed. It was NOT repointed to either Gaudi record because the referring context does not choose between them: all three of these records are sourced to the vLLM Gaudi plugin README and plugin-system documentation (sources xpu-vllm-gaudi-readme, xpu-vllm-gaudi-plugin-system), which are CURRENT-GENERATION-AGNOSTIC - the plugin supports Gaudi 2 and Gaudi 3 with the same hpu_* custom ops and the same HPU-graph mechanism on both. Choosing intel-gaudi3 would have been a guess dressed as a fix, and it would have added a SECOND Gaudi 3 reference to a field that already had one, implying a specificity the source does not have. The cost of deleting rather than repointing is nil in this field, because the coverage 'hpu' was standing in for - 'this applies to Gaudi hardware' - is already carried by the two sibling entries. The concept itself is NOT lost: it remains in gotchas xpu-gaudi-fork-after-driver-init and xpu-gaudi-fork-eol, whose concepts arrays carry 'hpu' as free-text scope, which is where a device-CLASS name belongs since concepts are explicitly not id-resolved. A future reader who wants the device-class grouping should look for it there, not in an id field that can only hold card ids.