qwen3-4b-gdn-hybrid-init
Surgery/init checkpoint — uniform 1:4 GDN conversion of Qwen3-4B with plain weight inheritance (99.88% of params inherited), before any training. Wikitext-2 PPL ~71,887 at init; kernel-level parity at the teacher noise floor (G1).
Part of a study converting full-attention Qwen3-4B into a GDN (gated
DeltaNet) hybrid (27 of 36 layers converted, uniform 1:4 retention) and
recovering capability via staged distillation, culminating in on-policy
distillation (OPD): student rollouts scored per-token by the frozen teacher
under reverse KL. Checkpoints are stock Qwen3NextForCausalLM — load with
AutoModelForCausalLM (transformers >= 4.57). vLLM serving needs a small
registry shim (all-dense MoE + tied-embeddings fixes) from the project repo.
Pipeline: surgery/init -> stage-1 hidden-state alignment (100M tok) -> stage-2a forward-KL KD @4k (600M) -> stage-2b long-context KD @32k (294M) -> stage-3 OPD (200M gen tokens, horizon 512->16k, token-keyed LR). Teacher/reference: Qwen/Qwen3-4B (thinking mode).