For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Mojo struct
Struct_ep_dispatch_block_scaled_nv_expert_load
struct Struct_ep_dispatch_block_scaled_nv_expert_load
Registers ep.dispatch.block.scaled.nv.expert_load: ep.dispatch.block.scaled.nv that also adds each local expert's routed-token count to an EPLB load tally.
Implemented traitsโ
Methodsโ
executeโ
static def execute[hidden_size: Int, top_k: Int, n_experts: Int, max_token_per_rank: Int, n_gpus_per_node: Int, n_nodes: Int, fused_shared_expert: Bool, skip_a2a: Bool, allreduce_world_size: Int, //](output_tokens: T, output_scales: T, row_offsets: T, scales_offsets: T, expert_ids: T, src_info: T, atomic_counters: T, input_tokens: T, topk_ids: T, send_ptrs: T, recv_ptrs: T, recv_count_ptrs: T, input_scales: T, expert_load: T, layer_idx: T, context: DeviceContext)
Runs ep.dispatch.block.scaled.nv and adds each local expert's routed-token count to row layer_idx[0] of the [layers, n_local_experts] uint64 expert_load tally.