arctic_inference/vllm/swiftkv/llama_swiftkv.py [606:612]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            for idx, layer in enumerate(
                    self.layers[self.config.num_key_value_layers:]):
                attn = layer.self_attn.attn
                kv_cache = attn.kv_cache[forward_context.virtual_engine]
                if kv_cache.numel():
                    # different cache layouts
                    if FLASHINFER_AVAILABLE and isinstance(attn_metadata, FlashInferMetadata):
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



arctic_inference/vllm/swiftkv/llama_swiftkv.py [635:640]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
            for idx, layer in enumerate(
                    self.layers[self.config.num_key_value_layers:]):
                attn = layer.self_attn.attn
                kv_cache = attn.kv_cache[forward_context.virtual_engine]
                if kv_cache.numel():
                    if FLASHINFER_AVAILABLE and isinstance(attn_metadata, FlashInferMetadata):
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -