- state.param_states.decoder.decoder_norm.scale.v
- state.param_states.decoder.layers_0.pre_cross_attention_layer_norm.scale.v
- state.param_states.decoder.layers_0.pre_mlp_layer_norm.scale.v
- state.param_states.decoder.layers_0.pre_self_attention_layer_norm.scale.v
- state.param_states.decoder.layers_1.pre_cross_attention_layer_norm.scale.v
- state.param_states.decoder.layers_1.pre_mlp_layer_norm.scale.v
- state.param_states.decoder.layers_1.pre_self_attention_layer_norm.scale.v
- state.param_states.decoder.layers_2.pre_cross_attention_layer_norm.scale.v
- state.param_states.decoder.layers_2.pre_mlp_layer_norm.scale.v
- state.param_states.decoder.layers_2.pre_self_attention_layer_norm.scale.v
- state.param_states.decoder.layers_3.pre_cross_attention_layer_norm.scale.v
- state.param_states.decoder.layers_3.pre_mlp_layer_norm.scale.v
- state.param_states.decoder.layers_3.pre_self_attention_layer_norm.scale.v
- state.param_states.decoder.layers_4.pre_cross_attention_layer_norm.scale.v
- state.param_states.decoder.layers_4.pre_mlp_layer_norm.scale.v
- state.param_states.decoder.layers_4.pre_self_attention_layer_norm.scale.v
- state.param_states.decoder.layers_5.pre_cross_attention_layer_norm.scale.v
- state.param_states.decoder.layers_5.pre_mlp_layer_norm.scale.v
- state.param_states.decoder.layers_5.pre_self_attention_layer_norm.scale.v
- state.param_states.decoder.relpos_bias.rel_embedding.v
- state.param_states.encoder.encoder_norm.scale.v
- state.param_states.encoder.layers_0.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_0.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_1.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_1.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_10.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_10.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_11.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_11.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_12.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_12.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_13.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_13.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_14.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_14.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_15.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_15.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_16.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_16.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_17.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_17.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_2.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_2.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_3.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_3.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_4.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_4.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_5.pre_attention_layer_norm.scale.v
- state.param_states.encoder.layers_5.pre_mlp_layer_norm.scale.v
- state.param_states.encoder.layers_6.pre_attention_layer_norm.scale.v