qwen3

qwen3

Qwen3 (Qwen3ForCausalLM).

Llama's tree and Llama's block: model.{embed_tokens, layers[i].{input_layernorm, self_attn, post_attention_layernorm, mlp}, norm} and lm_head, the residual added in the block, attention through the shared eager forward. q_norm / k_norm sit inside the attention; head_dim is not hidden // heads.

RENAME module-attribute

RENAME = {'model.embed_tokens': 'embed_tokens', 'model.layers': 'layers', 'model.norm': 'norm', 'model.language_model.embed_tokens': 'embed_tokens', 'model.language_model.layers': 'layers', 'model.language_model.norm': 'norm'}

ENVOYS module-attribute

ENVOYS = {Qwen3DecoderLayer: Layer, Qwen3Attention: Attention, Qwen3MLP: Mlp}