Skip to content

You are viewing the latest developer preview docs. Click here to view docs for the latest stable release.

vLLM

vllm.attention.ops.blocksparse_attention

Initializing search

Home
User Guide
Developer Guide
API Reference
CLI Reference
Community

vLLM

Home
User Guide
Developer Guide
API Reference
API Reference
- API Reference
- Contents
  Contents
  - vllm.beam_search
  - vllm.collect_env
  - vllm.config
  - vllm.connections
  - vllm.env_override
  - vllm.envs
  - vllm.forward_context
  - vllm
  - vllm.jsontree
  - vllm.logger
  - vllm.logits_process
  - vllm.outputs
  - vllm.pooling_params
  - vllm.sampling_params
  - vllm.scalar_type
  - vllm.scripts
  - vllm.sequence
  - vllm.test_utils
  - vllm.tracing
  - vllm.version
  - adapter_commons
  - assets
  - attention
    attention
    
    vllm.attention
    
    vllm.attention.layer
    
    vllm.attention.selector
    
    backends
    
    ops
    ops
    
    vllm.attention.ops
    
    vllm.attention.ops.chunked_prefill_paged_decode
    
    vllm.attention.ops.flashmla
    
    vllm.attention.ops.hpu_paged_attn
    
    vllm.attention.ops.ipex_attn
    
    vllm.attention.ops.merge_attn_states
    
    vllm.attention.ops.nki_flash_attn
    
    vllm.attention.ops.paged_attn
    
    vllm.attention.ops.pallas_kv_cache_update
    
    vllm.attention.ops.prefix_prefill
    
    vllm.attention.ops.rocm_aiter_mla
    
    vllm.attention.ops.rocm_aiter_paged_attn
    
    vllm.attention.ops.triton_decode_attention
    
    vllm.attention.ops.triton_flash_attention
    
    vllm.attention.ops.triton_merge_attn_states
    
    vllm.attention.ops.triton_unified_attention
    
    blocksparse_attention
    blocksparse_attention
    
    vllm.attention.ops.blocksparse_attention vllm.attention.ops.blocksparse_attention
    Table of contents
    
    blocksparse_attention
    
    vllm.attention.ops.blocksparse_attention.blocksparse_attention_kernel
    
    vllm.attention.ops.blocksparse_attention.interface
    
    vllm.attention.ops.blocksparse_attention.utils
  - benchmarks
  - compilation
  - core
  - device_allocator
  - distributed
  - engine
  - entrypoints
  - executor
  - inputs
  - logging_utils
  - lora
  - model_executor
  - multimodal
  - platforms
  - plugins
  - profiler
  - prompt_adapter
  - reasoning
  - spec_decode
  - transformers_utils
  - triton_utils
  - usage
  - utils
  - v1
  - worker
CLI Reference
Community

Table of contents

blocksparse_attention

vllm.attention.ops.blocksparse_attention

Modules:

Name	Description
`blocksparse_attention_kernel`
`interface`
`utils`

Made with Material for MkDocs