Skip to content

Qwen3 / other-model application contributions & decode-attention operator roadmap #156

Description

@ZhongYic00

Is your feature request related to a problem? Please describe.

We adapted Qwen3-0.6B to run end-to-end on NPU2 (Strix Point) with IRON. Two questions before we contribute:

  1. Does upstream accept model applications beyond Llama (e.g. Qwen3)? What are the acceptance criteria (app structure, test/benchmark requirements, weight handling)?
  2. Are there plans for first-class decode-oriented attention (fused GQA with runtime seq_pos, KV cache streamed from DRAM)? Today a model port must assemble decode attention from GEMV + Softmax + Transpose per layer.

Describe the solution you'd like

  • A place (or documented criteria) for model applications beyond Llama
  • A decode-attention operator (fused scores-GEMV + online-softmax + context-GEMV) as a standard operator

For reference, we have a working pure-IRON Qwen3-0.6B decode (28 layers in a single OperatorSequence, ~65 ms/token, logits cosine 0.9985 vs HuggingFace):
https://github.com/ZhongYic00/IRON/tree/feat/qwen3-decode

  • app: iron/applications/qwen3_0.6b/ (dialogue runner, HF cosine/e2e checks, TPOT bench, README)
  • new operators: decode_attn (fused decode attention, runtime seq_pos, S_KV up to 4096), qk_norm, gemv_argmax(_bf16), etc.

Describe alternatives you've considered

Hand-rolling per-model decode attention from GEMV + Softmax + Transpose (what the Llama app does). Works, but it is ~10 ops per layer with per-op array reconfiguration; a fused decode_attn operator collapses that to one.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions