
    #;7j                   X    d dl mZ d dlmZ ddlmZmZmZmZ  G d d          Z	ddZ
dS )    )annotations)Sequence   )
MODEL_ARCHMODEL_TENSORMODEL_TENSORSTENSOR_NAMESc                     e Zd ZU i ej        dej        dej        dej        dej        dej	        dej
        dej        dej        d	ej        d
ej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej         dej!        diZ"de#d<   i ej$        d ej%        d!ej&        d"ej'        d#ej(        d$ej)        d%ej*        d&ej+        d'ej,        d(ej-        d)ej.        d*ej/        d+ej0        d,ej1        d-ej2        d.ej3        d/ej4        d0i ej5        d1ej6        d2ej7        d3ej8        d4ej9        d5ej:        d6ej;        d7ej<        d8ej=        d9ej>        d:ej?        d;ej@        d<ejA        d=ejB        d>ejC        d?ejD        d@ejE        dAi ejF        dBejG        dCejH        dDejI        dEejJ        dFej        dGejK        dHejL        dIejM        dJejN        dKejO        dLejP        dMejQ        dNejR        dOejS        dPejT        dQejU        dRi ejV        dSejW        dTejX        dUejY        dVejZ        dWej[        dXej\        dYej]        dZej^        d[ej_        d\ej`        d]eja        d^ejb        d_ejc        d`ejd        daeje        dbejf        dci ejg        ddejh        deeji        dfejj        dgejk        dhejl        diejm        djejn        dkejo        dlejp        dmejq        dnejr        doejs        dpejt        dqeju        drejv        dsejw        dti ejx        duejy        dvejz        dwej{        dxej|        dyej}        dzej~        d{ej        d|ej        d}ej        d~ej        dej        dej        dej        dej        dej        dej        di ej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        di ej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        di ej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        di ej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dÓej        dēej        dœej        dƓej        dǓej        dȓej        dɓi ej        dʓej        d˓ej        d̓ej        d͓ej        d˓ej        dΓej        dϓej        dГej        dѓej        dғej        dӓej        dԓej        dՓej        d֓ej        dדej        dؓej        dٓi ej        dړej        dۓej        dܓej        dݓej        dޓej        dߓej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        di ej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        di ej        dej        dej         dej        dej        d ej        dej        dej        dej        dej        dej        dej	        dej
        dej        d	ej        d
ej        dej        di ej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        dej        di ej         dej!        dej"        d ej#        d!ej$        d"ej%        d#ej&        d$ej'        d%ej(        d&ej)        d'ej*        d(ej+        d)ej,        d*ej-        d+ej.        d,ej/        d-ej0        d.i ej1        d/ej2        d0ej3        d1ej4        d2ej5        d3ej6        d4ej7        d5ej8        d6ej9        d7ej:        d8ej;        d9ej<        d:ej=        d;ej>        d<ej?        d=ej@        d>ejA        d?i ejB        d@ejC        dAejD        dBejE        dCejF        dejG        dDejH        dEejI        dFejJ        dGejK        dHejL        dIejM        dJejN        dKejO        dLejP        dMejQ        dNejR        dOi ejS        dPejT        dQejU        dRejV        dSejW        dejX        dTejY        dUejZ        dVej[        dWej\        dXej]        dYej^        dZej_        d[ej`        d\eja        d]ejb        d^ejc        d_i ejd        d`eje        daejf        dbejg        dcejh        ddeji        deejj        dfejk        dgejl        dhejm        diejn        djejo        dkejp        dlejq        dmejr        dnejs        doejt        dpi eju        dqejv        drejw        dsejx        dtejy        duejz        dvej{        dwej|        dxej}        dyej~        dzej        d{ej        d|ej        d}ej        d~ej        dej        dej        dej        diZde#d<   ej        ej0        dej        diiZde#d<   de#d<   ddZdddZdddZdddZddZddZddZdS (  TensorNameMap)zgpt_neox.embed_inztransformer.wteztransformer.word_embeddingsword_embeddingszmodel.embed_tokensembed_tokenstok_embeddingszembeddings.word_embeddingszembeddings.tok_embeddingswteztransformer.embd.wtezmodel.tok_embeddingszmodel.embeddingzbackbone.embeddingzbackbone.embeddingsztransformer.in_out_embedzembedding.word_embeddingsztransformer.token_embeddingssharedzrwkv.embeddingszmodel.embeddingszmodel.word_embeddingsencoderzmodel.transformer.wter   zmodel.embed)zmasked_embedding.centroids)zmasked_embedding.token_ordering)z embeddings.token_type_embeddings)word_embeddings_layernormzembeddings.LayerNormzembeddings.normemb_lntransformer.normrwkv.blocks.0.pre_lnr   zmodel.pre_lnzmodel.layers.0.pre_normzbackbone.normzmodel.embedding_norm)ztransformer.wpezembeddings.position_embeddingswpezmodel.embed_positions)	embed_outlm_headoutputword_embeddings_for_headzlm_head.linearoutput_layerheadzhead.outr   zmodel.transformer.ff_outzhead.decoder)dense_2_out)dense_3_out)zgpt_neox.final_layer_normztransformer.ln_f
model.normnormztransformer.norm_fln_fzmodel.final_layernormz
lm_head.lnzmodel.norm_fzbackbone.norm_fztransformer.rms_normzencoder.final_layernormr   r   zrwkv.ln_outzmodel.ln_outzbackbone.final_layer_normr   zmodel.transformer.ln_f
final_normr   )z
rope.freqszrotary_pos_emb.inv_freq )zbackbone.embed)zmodel.embed_vision.embedding)z&model.embed_vision.hard_embedding_norm)z'model.embed_vision.embedding_projection)z&model.embed_vision.soft_embedding_norm)z,model.vision_tower.timm_model.conv_stem.conv)z*model.vision_tower.timm_model.conv_stem.bn)z2model.vision_tower.timm_model.msfa.ffn.pw_exp.conv)z0model.vision_tower.timm_model.msfa.ffn.pw_exp.bn)z3model.vision_tower.timm_model.msfa.ffn.pw_proj.conv)z1model.vision_tower.timm_model.msfa.ffn.pw_proj.bn)z'model.vision_tower.timm_model.msfa.norm)zencoder.out)zencoder.out_mid)zprojector.query)zprojector.qformer.layernorm)zprojector.linearz#dict[MODEL_TENSOR, tuple[str, ...]]mappings_cfg) z%gpt_neox.layers.{bid}.input_layernormztransformer.h.{bid}.ln_1ztransformer.blocks.{bid}.norm_1z#transformer.h.{bid}.input_layernormzh.{bid}.input_layernormztransformer.h.{bid}.ln_mlp"model.layers.{bid}.input_layernormzlayers.{bid}.attention_normmodel.layers.{bid}.ln1zh.{bid}.ln_1ztransformer.h.{bid}.lnzmodel.layers.layers.{bid}.normz(model.layers.layers.{bid}.pre_mixer_normz!model.layers.{bid}.attention_normzmodel.layers.{bid}.normzbackbone.layers.{bid}.normz(transformer.decoder_layer.{bid}.rms_normz model.layers.{bid}.pre_attn_normz.transformer.blocks.{bid}.norm_attn_norm.norm_1z$encoder.layers.{bid}.input_layernormz"transformer.layers.{bid}.attn_normzrwkv.blocks.{bid}.ln1r&   r%   layers.{bid}.input_layernormz(transformer_encoder.{bid}.attention_normzlayers.{bid}.attn_normz model.layers.{bid}.operator_normz(model.transformer.blocks.{bid}.attn_normr'   z&model.layers.{bid}.attention_layernormz*model.layers.{bid}.pre_attention_layernorm)ztransformer.h.{bid}.ln_attnz encoder.layer.{bid}.layer_norm_1zrwkv.blocks.{bid}.ln2model.layers.{bid}.ln2+model.layers.{bid}.post_attention_layernorm)z/gpt_neox.layers.{bid}.attention.query_key_valueztransformer.h.{bid}.attn.c_attnz"transformer.blocks.{bid}.attn.Wqkvz1transformer.blocks.{bid}.norm_attn_norm.attn.Wqkvz2transformer.h.{bid}.self_attention.query_key_valuez&h.{bid}.self_attention.query_key_valuez,model.layers.{bid}.self_attn.query_key_valuez,model.layers.{bid}.attention.query_key_valuezh.{bid}.attn.c_attnztransformer.h.{bid}.mixer.Wqkvzencoder.layers.{bid}.attn.Wqkvzencoder.layers.{bid}.mixer.Wqkvz%model.layers.{bid}.self_attn.qkv_projz(model.layers.layers.{bid}.mixer.qkv_projz3encoder.layers.{bid}.self_attention.query_key_valuez&transformer.layers.{bid}.attn.qkv_projztransformer_encoder.{bid}.qkvzlayers.{bid}.attn.Wqkvz<model.layers.{bid}.self_attn.language_expert_query_key_valuez*model.layers.{bid}.linear_attn.in_proj_qkv)#model.layers.{bid}.self_attn.q_projlayers.{bid}.self_attn.q_projz+model.layers.{bid}.self_attn.q_proj_no_permzlayers.{bid}.attention.wqz(encoder.layer.{bid}.attention.self.queryz'transformer.layer.{bid}.attention.q_linztransformer.h.{bid}.attn.q_projz*model.layers.layers.{bid}.self_attn.q_projzmodel.layers.{bid}.attention.wqz:transformer.decoder_layer.{bid}.multi_head_attention.queryz)transformer.h.{bid}.attn.attention.q_projr*   z%model.transformer.blocks.{bid}.q_projr+   z"backbone.layers.{bid}.mixer.q_projz"model.blocks.{bid}.attn.attn_query)#model.layers.{bid}.self_attn.k_projlayers.{bid}.self_attn.k_projz+model.layers.{bid}.self_attn.k_proj_no_permzlayers.{bid}.attention.wkz&encoder.layer.{bid}.attention.self.keyz'transformer.layer.{bid}.attention.k_linztransformer.h.{bid}.attn.k_projztransformer.h.{bid}.attn.kz*model.layers.layers.{bid}.self_attn.k_projzmodel.layers.{bid}.attention.wkz8transformer.decoder_layer.{bid}.multi_head_attention.keyz)transformer.h.{bid}.attn.attention.k_projr,   z%model.transformer.blocks.{bid}.k_projr-   z"backbone.layers.{bid}.mixer.k_projz model.blocks.{bid}.attn.attn_key)#model.layers.{bid}.self_attn.v_projlayers.{bid}.self_attn.v_projzlayers.{bid}.attention.wvz(encoder.layer.{bid}.attention.self.valuez'transformer.layer.{bid}.attention.v_linztransformer.h.{bid}.attn.v_projztransformer.h.{bid}.attn.vz*model.layers.layers.{bid}.self_attn.v_projzmodel.layers.{bid}.attention.wvz:transformer.decoder_layer.{bid}.multi_head_attention.valuez)transformer.h.{bid}.attn.attention.v_projr.   z%model.transformer.blocks.{bid}.v_projr/   z"backbone.layers.{bid}.mixer.v_projz"model.blocks.{bid}.attn.attn_value)#z%gpt_neox.layers.{bid}.attention.denseztransformer.h.{bid}.attn.c_projz&transformer.blocks.{bid}.attn.out_projz(transformer.h.{bid}.self_attention.densezh.{bid}.self_attention.dense#model.layers.{bid}.self_attn.o_projlayers.{bid}.self_attn.o_projz%model.layers.{bid}.self_attn.out_projz(model.layers.{bid}.self_attn.linear_attnzlayers.{bid}.attention.woz*encoder.layer.{bid}.attention.output.densezlayers.{bid}.attn.Woz)transformer.layer.{bid}.attention.out_linz!transformer.h.{bid}.attn.out_projz"model.layers.{bid}.self_attn.densez"model.layers.{bid}.attention.densezh.{bid}.attn.c_projz"transformer.h.{bid}.mixer.out_projz*model.layers.layers.{bid}.self_attn.o_projz&model.layers.layers.{bid}.mixer.o_projzmodel.layers.{bid}.attention.woz"encoder.layers.{bid}.attn.out_projz#encoder.layers.{bid}.mixer.out_projz;transformer.decoder_layer.{bid}.multi_head_attention.linearz5transformer.blocks.{bid}.norm_attn_norm.attn.out_projz)encoder.layers.{bid}.self_attention.densez&transformer.layers.{bid}.attn.out_projz+transformer.h.{bid}.attn.attention.out_projr0   ztransformer_encoder.{bid}.woz'model.transformer.blocks.{bid}.attn_outr1   z"backbone.layers.{bid}.mixer.o_projz2model.layers.{bid}.self_attn.language_expert_densez"model.blocks.{bid}.attn.attn_resid)z.encoder.layer.{bid}.attention.output.LayerNormz%transformer.layer.{bid}.sa_layer_normzencoder.layers.{bid}.norm1z*transformer.decoder_layer.{bid}.rms_norm_1z!model.layers.{bid}.post_attn_normz.transformer.blocks.{bid}.norm_attn_norm.norm_2)r)   %layers.{bid}.post_attention_layernormz+model.layers.{bid}.post_self_attn_layernormz0model.layers.layers.{bid}.post_mixer_norm.weight)z0model.layers.{bid}.self_attn.rotary_emb.inv_freqz1layers.{bid}.attention.inner_attention.rope.freqsz7model.layers.layers.{bid}.self_attn.rotary_emb.inv_freqz,transformer.h.{bid}.attn.rotary_emb.inv_freq)z"model.layers.{bid}.self_attn.sinksz0model.layers.{bid}.self_attn.attention_sink_bias)z&model.layers.{bid}.self_attn.gate_projz(model.layers.{bid}.linear_attn.in_proj_zz#model.layers.{bid}.self_attn.g_proj)z.gpt_neox.layers.{bid}.post_attention_layernormztransformer.h.{bid}.ln_2z h.{bid}.post_attention_layernormztransformer.blocks.{bid}.norm_2r)   zlayers.{bid}.ffn_normr(   zh.{bid}.ln_2zmodel.layers.{bid}.ffn_normz*transformer.decoder_layer.{bid}.rms_norm_2zmodel.layers.{bid}.pre_moe_normz-encoder.layers.{bid}.post_attention_layernormz!transformer.layers.{bid}.ffn_normz#model.layers.{bid}.pre_ff_layernormz$model.layers.{bid}.pre_moe_layernormr)   z"transformer_encoder.{bid}.ffn_normz&model.layers.layers.{bid}.pre_mlp_normz&model.transformer.blocks.{bid}.ff_normr2   z(model.layers.{bid}.feedforward_layernorm$model.layers.{bid}.pre_mlp_layernormzlayers.{bid}.mlp_norm)z,model.layers.{bid}.pre_feedforward_layernormz&layers.{bid}.pre_feedforward_layernormz*model.layers.{bid}.pre_ff_layernorm.weightr3   )z.model.layers.{bid}.pre_feedforward_layernorm_2)z-model.layers.{bid}.post_feedforward_layernormz'layers.{bid}.post_feedforward_layernormz%model.layers.{bid}.post_mlp_layernormz.model.layers.layers.{bid}.post_mlp_norm.weight'model.layers.{bid}.feed_forward.up_projz model.layers.{bid}.post_moe_norm)z/model.layers.{bid}.post_feedforward_layernorm_1)z/model.layers.{bid}.post_feedforward_layernorm_2)zlayers.{bid}.feed_forward.gatez(model.layers.{bid}.block_sparse_moe.gatezmodel.layers.{bid}.mlp.gatez&transformer.decoder_layer.{bid}.routerz)transformer.blocks.{bid}.ffn.router.layerz0model.layers.{bid}.block_sparse_moe.router.layerz&model.layers.{bid}.feed_forward.routerz%encoder.layers.{bid}.mlp.router.layerzmodel.layers.{bid}.mlp.routerzmodel.layers.{bid}.mlp.gate.wgz2model.layers.{bid}.block_sparse_moe.primary_routerz$model.layers.{bid}.feed_forward.gatez"model.layers.{bid}.mlp.router.gatezlayers.{bid}.gatez backbone.layers.{bid}.mixer.gatezmodel.layers.{bid}.moe.gatezmodel.layers.{bid}.router.proj)z)model.layers.{bid}.mlp.shared_expert_gate)
z.model.layers.{bid}.mlp.gate.e_score_correctionz5model.layers.{bid}.mlp.moe_statics.e_score_correctionz'model.layers.{bid}.mlp.gate.expert_biasz"model.layers.{bid}.mlp.expert_biasz+model.layers.{bid}.feed_forward.expert_biasz6model.layers.{bid}.block_sparse_moe.e_score_correctionz3backbone.layers.{bid}.mixer.gate.e_score_correctionz)model.layers.{bid}.mlp.e_score_correctionz;model.layers.{bid}.block_sparse_moe.gate.e_score_correctionz"model.layers.{bid}.moe.router_bias)'z'gpt_neox.layers.{bid}.mlp.dense_h_to_4hztransformer.h.{bid}.mlp.c_fcz$transformer.blocks.{bid}.ffn.up_projz%transformer.h.{bid}.mlp.dense_h_to_4hzh.{bid}.mlp.dense_h_to_4hzmodel.layers.{bid}.mlp.up_projlayers.{bid}.mlp.up_projzlayers.{bid}.feed_forward.w3z&encoder.layer.{bid}.intermediate.densezlayers.{bid}.mlp.Wiz transformer.layer.{bid}.ffn.lin1ztransformer.h.{bid}.mlp.fc_inz transformer.h.{bid}.mlp.linear_3z$model.layers.{bid}.mlp.dense_h_to_4hztransformer.h.{bid}.mlp.w1zh.{bid}.mlp.c_fcztransformer.h.{bid}.mlp.fc1zmodel.layers.{bid}.mlp.fc1z#model.layers.{bid}.mlp.gate_up_projz%model.layers.layers.{bid}.mlp.up_projz*model.layers.layers.{bid}.mlp.gate_up_projz"model.layers.{bid}.feed_forward.w3zencoder.layers.{bid}.mlp.fc11zencoder.layers.{bid}.mlp.fc1zmodel.layers.{bid}.mlp.c_fcz&encoder.layer.{bid}.mlp.gated_layers_vz$encoder.layer.{bid}.mlp.gated_layersz&encoder.layer.{bid}.mlp.up_gated_layerz"model.layers.{bid}.residual_mlp.w3z&encoder.layers.{bid}.mlp.dense_h_to_4hztransformer.h.{bid}.mlp.c_fc_1r4   z!transformer_encoder.{bid}.ffn.w12z&model.layers.{bid}.block_sparse_moe.upz&model.transformer.blocks.{bid}.up_projr5   z#backbone.layers.{bid}.mixer.up_projz+model.layers.{bid}.mlp.language_mlp.up_projz!model.blocks.{bid}.mlp.mlp_linear)	z$layers.{bid}.feed_forward.experts.w3z,transformer.decoder_layer.{bid}.moe.linear_vz+transformer.blocks.{bid}.ffn.experts.mlp.v1z&model.layers.{bid}.mlp.experts.up_projz.model.layers.{bid}.block_sparse_moe.experts.w3z/model.layers.{bid}.feed_forward.experts.up_projz'encoder.layers.{bid}.mlp.experts.mlp.w1z.model.layers.{bid}.block_sparse_moe.experts.upzmodel.layers.{bid}.moe.up_proj)	z,model.layers.{bid}.mlp.shared_expert.up_projz-model.layers.{bid}.mlp.shared_experts.up_projz5model.layers.{bid}.feed_forward.shared_expert.up_proj)model.layers.{bid}.feed_forward.down_projz)model.layers.{bid}.mlp.shared_mlp.up_projzlayers.{bid}.shared_experts.w3z2backbone.layers.{bid}.mixer.shared_experts.up_projz:model.layers.{bid}.block_sparse_moe.shared_experts.up_projz'model.layers.{bid}.share_expert.up_proj)z,model.layers.{bid}.mlp.chunk_experts.up_proj)z transformer.blocks.{bid}.ffn.act)z model.layers.{bid}.mlp.gate_projlayers.{bid}.mlp.gate_projzlayers.{bid}.feed_forward.w1ztransformer.h.{bid}.mlp.w2ztransformer.h.{bid}.mlp.c_fc2z'model.layers.layers.{bid}.mlp.gate_projz"model.layers.{bid}.feed_forward.w1zencoder.layers.{bid}.mlp.fc12z&encoder.layer.{bid}.mlp.gated_layers_wz transformer.h.{bid}.mlp.linear_1z"model.layers.{bid}.residual_mlp.w1ztransformer.h.{bid}.mlp.c_fc_0z)model.layers.{bid}.feed_forward.gate_projz&model.transformer.blocks.{bid}.ff_projr7   z-model.layers.{bid}.mlp.language_mlp.gate_projzmodel.blocks.{bid}.mlp.mlp_gate)z$layers.{bid}.feed_forward.experts.w1z*transformer.decoder_layer.{bid}.moe.linearz+transformer.blocks.{bid}.ffn.experts.mlp.w1z(model.layers.{bid}.mlp.experts.gate_projz.model.layers.{bid}.block_sparse_moe.experts.w1z1model.layers.{bid}.feed_forward.experts.gate_projz0model.layers.{bid}.block_sparse_moe.experts.gatez model.layers.{bid}.moe.gate_proj)z.model.layers.{bid}.mlp.shared_expert.gate_projz/model.layers.{bid}.mlp.shared_experts.gate_projz7model.layers.{bid}.feed_forward.shared_expert.gate_projz+model.layers.{bid}.mlp.shared_mlp.gate_projzlayers.{bid}.shared_experts.w1z<model.layers.{bid}.block_sparse_moe.shared_experts.gate_projz)model.layers.{bid}.share_expert.gate_proj)z.model.layers.{bid}.mlp.chunk_experts.gate_proj)z+model.layers.{bid}.mlp.experts.gate_up_projz'model.layers.{bid}.experts.gate_up_proj)z+backbone.layers.{bid}.mixer.fc1_latent_proj)z+backbone.layers.{bid}.mixer.fc2_latent_proj)"z'gpt_neox.layers.{bid}.mlp.dense_4h_to_hztransformer.h.{bid}.mlp.c_projz&transformer.blocks.{bid}.ffn.down_projz%transformer.h.{bid}.mlp.dense_4h_to_hzh.{bid}.mlp.dense_4h_to_hz model.layers.{bid}.mlp.down_projlayers.{bid}.mlp.down_projzlayers.{bid}.feed_forward.w2z encoder.layer.{bid}.output.densezlayers.{bid}.mlp.Woz transformer.layer.{bid}.ffn.lin2ztransformer.h.{bid}.mlp.fc_outz$model.layers.{bid}.mlp.dense_4h_to_hzh.{bid}.mlp.c_projztransformer.h.{bid}.mlp.fc2zmodel.layers.{bid}.mlp.fc2z'model.layers.layers.{bid}.mlp.down_projz"model.layers.{bid}.feed_forward.w2zencoder.layers.{bid}.mlp.fc2zmodel.layers.{bid}.mlp.c_projzencoder.layer.{bid}.mlp.woz#transformer.layers.{bid}.ffn.proj_2z"model.layers.{bid}.residual_mlp.w2z"encoder.layer.{bid}.mlp.down_layerz&encoder.layers.{bid}.mlp.dense_4h_to_hzmodel.layers.h.{bid}.mlp.c_projr6   z transformer_encoder.{bid}.ffn.w3z(model.layers.{bid}.block_sparse_moe.downz%model.transformer.blocks.{bid}.ff_outr8   z%backbone.layers.{bid}.mixer.down_projz-model.layers.{bid}.mlp.language_mlp.down_projz model.blocks.{bid}.mlp.mlp_resid)z$layers.{bid}.feed_forward.experts.w2z,transformer.decoder_layer.{bid}.moe.linear_1z+transformer.blocks.{bid}.ffn.experts.mlp.w2z(model.layers.{bid}.mlp.experts.down_projz1model.layers.{bid}.block_sparse_moe.output_linearz.model.layers.{bid}.block_sparse_moe.experts.w2z1model.layers.{bid}.feed_forward.experts.down_projz'encoder.layers.{bid}.mlp.experts.mlp.w2z0model.layers.{bid}.block_sparse_moe.experts.downz model.layers.{bid}.moe.down_projz$model.layers.{bid}.experts.down_proj)	z.model.layers.{bid}.mlp.shared_expert.down_projz/model.layers.{bid}.mlp.shared_experts.down_projz7model.layers.{bid}.feed_forward.shared_expert.down_projz+model.layers.{bid}.shared_mlp.output_linearz+model.layers.{bid}.mlp.shared_mlp.down_projzlayers.{bid}.shared_experts.w2z4backbone.layers.{bid}.mixer.shared_experts.down_projz<model.layers.{bid}.block_sparse_moe.shared_experts.down_projz)model.layers.{bid}.share_expert.down_proj)z.model.layers.{bid}.mlp.chunk_experts.down_proj)z/encoder.layers.{bid}.self_attention.q_layernormz(model.layers.{bid}.self_attn.q_layernormz,model.layers.{bid}.self_attn.query_layernorm,model.layers.{bid}.attention.query_layernormz#model.layers.{bid}.self_attn.q_normlayers.{bid}.self_attn.q_normz"transformer.blocks.{bid}.attn.q_lnz/encoder.layer.{bid}.attention.self.layer_norm_qz$transformer.layers.{bid}.attn.q_normz!model.layers.layers.{bid}.mixer.qz&model.layers.layers.{bid}.mixer.q_normr:   r9   z(model.blocks.{bid}.attn.head_gain.head_g)z/encoder.layers.{bid}.self_attention.k_layernormz(model.layers.{bid}.self_attn.k_layernormz*model.layers.{bid}.self_attn.key_layernorm*model.layers.{bid}.attention.key_layernormz#model.layers.{bid}.self_attn.k_normlayers.{bid}.self_attn.k_normz"transformer.blocks.{bid}.attn.k_lnz/encoder.layer.{bid}.attention.self.layer_norm_kz$transformer.layers.{bid}.attn.k_normz!model.layers.layers.{bid}.mixer.kz&model.layers.layers.{bid}.mixer.k_normr<   r;   )z7encoder.layers.{bid}.self_attention.rotary_emb.inv_freq)z$encoder.layer.{bid}.output.LayerNormz)transformer.layer.{bid}.output_layer_normzencoder.layers.{bid}.norm2z*transformer.decoder_layer.{bid}.rms_norm_3z!encoder.layer.{bid}.mlp.layernormz encoder.layer.{bid}.layer_norm_2z"model.layers.{bid}.final_layernorm)zmodel.layers.{bid}.layer_scalarz!model.blocks.{bid}.embed_skip.a_g)zmodel.embed_tokens_per_layer)z model.per_layer_model_projection)zmodel.per_layer_projection_norm)zmodel.altup_projections)zmodel.altup_unembed_projections)z'model.layers.{bid}.per_layer_input_gate)z'model.layers.{bid}.per_layer_projection)z,model.layers.{bid}.post_per_layer_input_norm)z)model.layers.{bid}.altup.correction_coefs)z-model.layers.{bid}.altup.correct_output_scale)z)model.layers.{bid}.altup.prediction_coefs)z(model.layers.{bid}.altup.modality_router)z$model.layers.{bid}.altup.router_norm)z%model.layers.{bid}.laurel.linear_left)z&model.layers.{bid}.laurel.linear_right)z*model.layers.{bid}.laurel.post_laurel_norm)zmodel.layers.{bid}.in_projz#backbone.layers.{bid}.mixer.in_projz model.layers.{bid}.mamba.in_projz'model.layers.layers.{bid}.mixer.in_projz+model.layers.{bid}.linear_attn.in_proj_qkvz)zmodel.layers.{bid}.conv1dz"backbone.layers.{bid}.mixer.conv1dzmodel.layers.{bid}.mamba.conv1dz&model.layers.layers.{bid}.mixer.conv1dz%model.layers.{bid}.linear_attn.conv1d)zmodel.layers.{bid}.x_projz"backbone.layers.{bid}.mixer.x_projzmodel.layers.{bid}.mamba.x_projz)model.layers.layers.{bid}.mixer.bcdt_proj)zmodel.layers.{bid}.dt_projz#backbone.layers.{bid}.mixer.dt_projz model.layers.{bid}.mamba.dt_projz'model.layers.layers.{bid}.mixer.dt_projz&model.layers.{bid}.linear_attn.dt_projzbackbone.layers.{bid}.mixer.dtz$model.layers.{bid}.self_attn.dt_proj)z.model.layers.layers.{bid}.mixer.dt_norm.weightz%model.layers.{bid}.mamba.dt_layernorm)zmodel.layers.{bid}.A_logz!backbone.layers.{bid}.mixer.A_logzmodel.layers.{bid}.mamba.A_logz%model.layers.layers.{bid}.mixer.A_logz$model.layers.{bid}.linear_attn.A_logz"model.layers.{bid}.self_attn.A_log)z$model.layers.{bid}.mamba.b_layernormz$model.layers.{bid}.mamba.B_layernormz-model.layers.layers.{bid}.mixer.B_norm.weight)z$model.layers.{bid}.mamba.c_layernormz$model.layers.{bid}.mamba.C_layernormz-model.layers.layers.{bid}.mixer.C_norm.weight)zmodel.layers.{bid}.Dzbackbone.layers.{bid}.mixer.Dzmodel.layers.{bid}.mamba.Dz!model.layers.layers.{bid}.mixer.D)zmodel.layers.{bid}.mamba.normz#model.layers.{bid}.linear_attn.normz backbone.layers.{bid}.mixer.normz#model.layers.{bid}.self_attn.o_norm)zmodel.layers.{bid}.out_projz$backbone.layers.{bid}.mixer.out_projz!model.layers.{bid}.mamba.out_projz'model.layers.{bid}.linear_attn.out_projz(model.layers.layers.{bid}.mixer.out_proj)z(model.layers.{bid}.linear_attn.in_proj_a)z)model.layers.{bid}.linear_attn.in_proj_ba)z%model.layers.{bid}.self_attn.q_conv1d)z%model.layers.{bid}.self_attn.k_conv1d)z%model.layers.{bid}.self_attn.v_conv1d)z%model.layers.{bid}.self_attn.f_a_proj)z%model.layers.{bid}.self_attn.f_b_proj)z(model.layers.{bid}.linear_attn.in_proj_bz#model.layers.{bid}.self_attn.b_proj)z%model.layers.{bid}.self_attn.g_a_proj)z%model.layers.{bid}.self_attn.g_b_proj)zmodel.layers.{bid}.attention.w0)z'rwkv.blocks.{bid}.attention.time_maa_w1z(model.layers.{bid}.self_attn.time_maa_w1zmodel.layers.{bid}.attention.w1)z'rwkv.blocks.{bid}.attention.time_maa_w2z(model.layers.{bid}.self_attn.time_maa_w2zmodel.layers.{bid}.attention.w2)zmodel.layers.{bid}.attention.a0)zmodel.layers.{bid}.attention.a1)zmodel.layers.{bid}.attention.a2)zmodel.layers.{bid}.attention.v0)zmodel.layers.{bid}.attention.v1)zmodel.layers.{bid}.attention.v2)zmodel.layers.{bid}.attention.g1)zmodel.layers.{bid}.attention.g2)z model.layers.{bid}.attention.k_k)z model.layers.{bid}.attention.k_a)z model.layers.{bid}.attention.r_k)z&rwkv.blocks.{bid}.attention.time_maa_xz'model.layers.{bid}.self_attn.time_maa_x)z&rwkv.blocks.{bid}.attention.time_maa_kz'model.layers.{bid}.self_attn.time_maa_k)z&rwkv.blocks.{bid}.attention.time_maa_vz'model.layers.{bid}.self_attn.time_maa_v)z&rwkv.blocks.{bid}.attention.time_maa_rz'model.layers.{bid}.self_attn.time_maa_r)z&rwkv.blocks.{bid}.attention.time_maa_gz'model.layers.{bid}.self_attn.time_maa_g)z&rwkv.blocks.{bid}.attention.time_maa_wz'model.layers.{bid}.self_attn.time_maa_w)z&rwkv.blocks.{bid}.attention.time_faaaa)z&rwkv.blocks.{bid}.attention.time_decayz'model.layers.{bid}.self_attn.time_decay)z)rwkv.blocks.{bid}.attention.time_decay_w1z*model.layers.{bid}.self_attn.time_decay_w1)z)rwkv.blocks.{bid}.attention.time_decay_w2z*model.layers.{bid}.self_attn.time_decay_w2)zrwkv.blocks.{bid}.attention.keyr,   z model.layers.{bid}.attention.keyz#model.layers.{bid}.attention.k_proj)z!rwkv.blocks.{bid}.attention.valuer.   z"model.layers.{bid}.attention.valuez#model.layers.{bid}.attention.v_proj)z&rwkv.blocks.{bid}.attention.receptancer*   z'model.layers.{bid}.attention.receptancez#model.layers.{bid}.attention.r_proj)z rwkv.blocks.{bid}.attention.gatez!model.layers.{bid}.self_attn.gate)z rwkv.blocks.{bid}.attention.ln_xz!model.layers.{bid}.attention.ln_x)z"rwkv.blocks.{bid}.attention.outputr0   z#model.layers.{bid}.attention.outputz#model.layers.{bid}.attention.o_proj)z)rwkv.blocks.{bid}.feed_forward.time_maa_kz#model.layers.{bid}.feed_forward.x_k)z)rwkv.blocks.{bid}.feed_forward.time_maa_r)z"rwkv.blocks.{bid}.feed_forward.keyz#model.layers.{bid}.feed_forward.key)z)rwkv.blocks.{bid}.feed_forward.receptance)z$rwkv.blocks.{bid}.feed_forward.valuez%model.layers.{bid}.feed_forward.value)z%model.layers.{bid}.self_attn.q_a_projzlayers.{bid}.attention.wq_a)z%model.layers.{bid}.self_attn.q_b_projzlayers.{bid}.attention.wq_b)z/model.layers.{bid}.self_attn.kv_a_proj_with_mqaz%layers.{bid}.attention.wkv_a_with_mqa)z&model.layers.{bid}.self_attn.kv_b_proj)z%model.layers.{bid}.self_attn.k_b_projzlayers.{bid}.attention.k_b_proj)z%model.layers.{bid}.self_attn.v_b_projzlayers.{bid}.attention.v_b_proj)z*model.layers.{bid}.self_attn.q_a_layernormzlayers.{bid}.attention.q_a_norm)z+model.layers.{bid}.self_attn.kv_a_layernormz layers.{bid}.attention.kv_a_norm)z*model.layers.{bid}.self_attn.inner_attn_ln)z$model.layers.{bid}.mlp.ffn_layernorm)z&decoder.block.{bid}.layer.0.layer_norm)z+decoder.block.{bid}.layer.0.SelfAttention.q)z+decoder.block.{bid}.layer.0.SelfAttention.k)z+decoder.block.{bid}.layer.0.SelfAttention.v)z+decoder.block.{bid}.layer.0.SelfAttention.o)zAdecoder.block.{bid}.layer.0.SelfAttention.relative_attention_bias)z&decoder.block.{bid}.layer.1.layer_norm)z-decoder.block.{bid}.layer.1.EncDecAttention.q)z-decoder.block.{bid}.layer.1.EncDecAttention.k)z-decoder.block.{bid}.layer.1.EncDecAttention.v)z-decoder.block.{bid}.layer.1.EncDecAttention.o)zCdecoder.block.{bid}.layer.1.EncDecAttention.relative_attention_bias)z&decoder.block.{bid}.layer.2.layer_norm)z/decoder.block.{bid}.layer.2.DenseReluDense.wi_0)z-decoder.block.{bid}.layer.2.DenseReluDense.wiz/decoder.block.{bid}.layer.2.DenseReluDense.wi_1)z-decoder.block.{bid}.layer.2.DenseReluDense.wo)zdecoder.final_layer_norm)z&encoder.block.{bid}.layer.0.layer_norm)z+encoder.block.{bid}.layer.0.SelfAttention.q)z+encoder.block.{bid}.layer.0.SelfAttention.k)z+encoder.block.{bid}.layer.0.SelfAttention.v)z+encoder.block.{bid}.layer.0.SelfAttention.o)zAencoder.block.{bid}.layer.0.SelfAttention.relative_attention_bias)z&encoder.block.{bid}.layer.1.layer_norm)z/encoder.block.{bid}.layer.1.DenseReluDense.wi_0)z-encoder.block.{bid}.layer.1.DenseReluDense.wiz/encoder.block.{bid}.layer.1.DenseReluDense.wi_1)z-encoder.block.{bid}.layer.1.DenseReluDense.wo)z)model.layers.{bid}.mlp.vision_mlp.up_proj)z+model.layers.{bid}.mlp.vision_mlp.gate_proj)z+model.layers.{bid}.mlp.vision_mlp.down_proj)z0model.layers.{bid}.self_attn.vision_expert_dense)z:model.layers.{bid}.self_attn.vision_expert_query_key_value)z+model.layers.{bid}.self_attn.indexer.k_norm)z1model.layers.{bid}.self_attn.indexer.weights_proj)z'model.layers.{bid}.self_attn.indexer.wk)z)model.layers.{bid}.self_attn.indexer.wq_b)zencoder.final_layer_norm
layer_norm)
classifierzclassifier.densepre_classifierdensez
head.dense)zclassifier.out_proj)z	head.norm)zbackbone.convnext.{bid}.dwconv)zbackbone.convnext.{bid}.norm)zbackbone.convnext.{bid}.pwconv1)zbackbone.convnext.{bid}.pwconv2)zbackbone.convnext.{bid}.gamma)zbackbone.posnet.{bid}.conv1)zbackbone.posnet.{bid}.conv2)zbackbone.posnet.{bid}.norm)zbackbone.posnet.{bid}.norm1)zbackbone.posnet.{bid}.norm2)zbackbone.posnet.{bid}.q)zbackbone.posnet.{bid}.k)zbackbone.posnet.{bid}.v)zbackbone.posnet.{bid}.proj_out)zmodel.layers.{bid}.conv.conv)zmodel.layers.{bid}.conv.in_proj)z model.layers.{bid}.conv.out_proj)z"multi_modal_projector.linear_{bid}zmm_projector.proj.linear_{bid}zvisual.merger.mlp.{bid}zmlp_AR.linear_{bid}zmerger.mlp.{bid}zvision_tower.merger.mlp.{bid}zvit.perceive.proj.{bid})z(model.connector.modality_projection.projz$model.vision.linear_proj.linear_projzmodel.projector.layerszvisual.merger.projzvit.perceive.mlp)z model.mm_projector.mlp.mlp.{bid}z'vision_model.vision_adapter.mlp.fc{bid}z
mlp1.{bid}z%model.aligner.fc1.hidden_layers.{bid})z model.mm_projector.peg.peg.{bid})z4vision_tower.vision_model.embeddings.class_embeddingz'model.vision_tower.embeddings.cls_tokenzvision_model.class_embeddingz*model.vision.patch_embedding.cls_embeddingz>vision_model.radio_model.model.patch_generator.cls_token.tokenz-model.vision_model.embeddings.class_embedding)z:model.vision_tower.vision_model.embeddings.patch_embeddingz4vision_tower.vision_model.embeddings.patch_embeddingz-model.vision_tower.embeddings.patch_embeddingz9model.vision_tower.embeddings.patch_embeddings.projectionzvpm.embeddings.patch_embedding-model.vision_model.embeddings.patch_embeddingzvit.embeddings.patch_embeddingzvision_tower.patch_convzvision_encoder.patch_convz#vision_model.patch_embedding.linearzvisual.patch_embed.projzvision_tower.patch_embed.projz!model.vision.patch_embedding.projrA   z/siglip2.vision_model.embeddings.patch_embeddingz7vision_model.radio_model.model.patch_generator.embedderz,model.vision_tower.patch_embedder.input_projz(vision_tower.patch_embed.patchifier.projzvision_model.conv1z!model.vision_embedder.patch_dense)zvisual.post_conv_layernorm(vision_tower.patch_embed.patchifier.norm)z#model.vision_embedder.patch_ln{bid})z=model.vision_tower.vision_model.embeddings.position_embeddingz7vision_tower.vision_model.embeddings.position_embeddingz0model.vision_tower.embeddings.position_embeddingz1model.vision_tower.embeddings.position_embeddingsz!vpm.embeddings.position_embeddingz0model.vision_model.embeddings.position_embeddingz!vit.embeddings.position_embeddingz%vision_model.positional_embedding_vlmz vision_tower.patch_embed.pos_embzvisual.pos_embedz/model.vision.patch_embedding.position_embeddingz$visual.embeddings.position_embeddingz8vision_model.radio_model.model.patch_generator.pos_embedz:model.vision_tower.patch_embedder.position_embedding_tablez!vision_model.positional_embeddingz#model.vision_embedder.pos_embedding)zmodel.image_newlinezvit.perceive.image_newline)zmodel.view_seperatorzvit.perceive.image_sep)zvisual.blocks.{bid}.attn.qkvz"vision_tower.blocks.{bid}.attn.qkvz?model.vision.transformer.layers.{bid}.attention.query_key_valuez>model.vision_model.transformer.layers.{bid}.self_attn.qkv_projz&vision_tower.encoder.blocks.{bid}.wqkvz4vision_model.radio_model.model.blocks.{bid}.attn.qkvz5vision_model.transformer.resblocks.{bid}.attn.in_proj)zEmodel.vision_tower.vision_model.encoder.layers.{bid}.self_attn.q_projz?vision_tower.vision_model.encoder.layers.{bid}.self_attn.q_projz8model.vision_tower.encoder.layers.{bid}.self_attn.q_projz7model.vision_tower.encoder.layer.{bid}.attention.q_projz)vpm.encoder.layers.{bid}.self_attn.q_projz8model.vision_model.encoder.layers.{bid}.self_attn.q_projz!vit.layers.{bid}.self_attn.q_projz0vision_model.model.layers.{bid}.self_attn.q_projz6vision_tower.transformer.layers.{bid}.attention.q_projz4vision_encoder.transformer.layers.{bid}.attention.wqzvisual.blocks.{bid}.attn.qz$vision_tower.encoder.blocks.{bid}.wqz:siglip2.vision_model.encoder.layers.{bid}.self_attn.q_projz<model.vision_model.transformer.layers.{bid}.self_attn.q_projz7vision_model.model.layers.{bid}.self_attn.q_proj.linearz;model.qwen2_model.model.model.layers.{bid}.self_attn.q_proj)z:vision_tower.vision_model.encoder.layers.{bid}.attn.q_normz7model.vision_tower.encoder.layer.{bid}.attention.q_normzvisual.blocks.{bid}.attn.q_normz0vision_model.model.layers.{bid}.self_attn.q_norm)zEmodel.vision_tower.vision_model.encoder.layers.{bid}.self_attn.k_projz?vision_tower.vision_model.encoder.layers.{bid}.self_attn.k_projz8model.vision_tower.encoder.layers.{bid}.self_attn.k_projz7model.vision_tower.encoder.layer.{bid}.attention.k_projz)vpm.encoder.layers.{bid}.self_attn.k_projz8model.vision_model.encoder.layers.{bid}.self_attn.k_projz!vit.layers.{bid}.self_attn.k_projz0vision_model.model.layers.{bid}.self_attn.k_projz6vision_tower.transformer.layers.{bid}.attention.k_projz4vision_encoder.transformer.layers.{bid}.attention.wkzvisual.blocks.{bid}.attn.kz$vision_tower.encoder.blocks.{bid}.wkz<model.vision_model.transformer.layers.{bid}.self_attn.k_projz:siglip2.vision_model.encoder.layers.{bid}.self_attn.k_projz7vision_model.model.layers.{bid}.self_attn.k_proj.linearz;model.qwen2_model.model.model.layers.{bid}.self_attn.k_proj)z:vision_tower.vision_model.encoder.layers.{bid}.attn.k_normz7model.vision_tower.encoder.layer.{bid}.attention.k_normzvisual.blocks.{bid}.attn.k_normz0vision_model.model.layers.{bid}.self_attn.k_norm)zEmodel.vision_tower.vision_model.encoder.layers.{bid}.self_attn.v_projz?vision_tower.vision_model.encoder.layers.{bid}.self_attn.v_projz8model.vision_tower.encoder.layers.{bid}.self_attn.v_projz7model.vision_tower.encoder.layer.{bid}.attention.v_projz)vpm.encoder.layers.{bid}.self_attn.v_projz8model.vision_model.encoder.layers.{bid}.self_attn.v_projz!vit.layers.{bid}.self_attn.v_projz0vision_model.model.layers.{bid}.self_attn.v_projz6vision_tower.transformer.layers.{bid}.attention.v_projz4vision_encoder.transformer.layers.{bid}.attention.wvzvisual.blocks.{bid}.attn.vz$vision_tower.encoder.blocks.{bid}.wvz:siglip2.vision_model.encoder.layers.{bid}.self_attn.v_projz<model.vision_model.transformer.layers.{bid}.self_attn.v_projz7vision_model.model.layers.{bid}.self_attn.v_proj.linearz;model.qwen2_model.model.model.layers.{bid}.self_attn.v_proj)z@model.vision_tower.vision_model.encoder.layers.{bid}.layer_norm1z:vision_tower.vision_model.encoder.layers.{bid}.layer_norm1z3model.vision_tower.encoder.layers.{bid}.layer_norm1z4vision_tower.vision_model.encoder.layers.{bid}.norm1z7model.vision_tower.encoder.layer.{bid}.layernorm_beforez$vpm.encoder.layers.{bid}.layer_norm1z3model.vision_model.encoder.layers.{bid}.layer_norm1z vit.layers.{bid}.input_layernormz4vision_tower.transformer.layers.{bid}.attention_normz6vision_encoder.transformer.layers.{bid}.attention_normz/vision_model.model.layers.{bid}.input_layernormzvisual.blocks.{bid}.norm1z'vision_tower.encoder.blocks.{bid}.norm0z5model.vision.transformer.layers.{bid}.input_layernormz7model.vision_model.transformer.layers.{bid}.layer_norm1z5siglip2.vision_model.encoder.layers.{bid}.layer_norm1z1vision_model.radio_model.model.blocks.{bid}.norm1zvision_tower.blocks.{bid}.norm1z-vision_model.transformer.resblocks.{bid}.ln_1z:model.qwen2_model.model.model.layers.{bid}.input_layernorm)zGmodel.vision_tower.vision_model.encoder.layers.{bid}.self_attn.out_projzAvision_tower.vision_model.encoder.layers.{bid}.self_attn.out_projz:model.vision_tower.encoder.layers.{bid}.self_attn.out_projz8vision_tower.vision_model.encoder.layers.{bid}.attn.projzAmodel.vision_tower.encoder.layer.{bid}.attention.projection_layerz+vpm.encoder.layers.{bid}.self_attn.out_projz:model.vision_model.encoder.layers.{bid}.self_attn.out_projz!vit.layers.{bid}.self_attn.o_projzBmodel.vision_model.encoder.layers.{bid}.self_attn.projection_layerz0vision_model.model.layers.{bid}.self_attn.o_projz6vision_tower.transformer.layers.{bid}.attention.o_projz4vision_encoder.transformer.layers.{bid}.attention.wozvisual.blocks.{bid}.attn.projz$vision_tower.encoder.blocks.{bid}.woz5model.vision.transformer.layers.{bid}.attention.densez>model.vision_model.transformer.layers.{bid}.self_attn.out_projz<siglip2.vision_model.encoder.layers.{bid}.self_attn.out_projz5vision_model.radio_model.model.blocks.{bid}.attn.projz;model.qwen2_model.model.model.layers.{bid}.self_attn.o_projz7vision_model.model.layers.{bid}.self_attn.o_proj.linearz#vision_tower.blocks.{bid}.attn.projz6vision_model.transformer.resblocks.{bid}.attn.out_proj)zvisual.blocks.{bid}.attn.sinks)z@model.vision_tower.vision_model.encoder.layers.{bid}.layer_norm2z:vision_tower.vision_model.encoder.layers.{bid}.layer_norm2z3model.vision_tower.encoder.layers.{bid}.layer_norm2z4vision_tower.vision_model.encoder.layers.{bid}.norm2z6model.vision_tower.encoder.layer.{bid}.layernorm_afterz$vpm.encoder.layers.{bid}.layer_norm2z3model.vision_model.encoder.layers.{bid}.layer_norm2z)vit.layers.{bid}.post_attention_layernorm8vision_model.model.layers.{bid}.post_attention_layernormz.vision_tower.transformer.layers.{bid}.ffn_normz0vision_encoder.transformer.layers.{bid}.ffn_normzvisual.blocks.{bid}.norm2z'vision_tower.encoder.blocks.{bid}.norm1z>model.vision.transformer.layers.{bid}.post_attention_layernormz7model.vision_model.transformer.layers.{bid}.layer_norm2z5siglip2.vision_model.encoder.layers.{bid}.layer_norm2z1vision_model.radio_model.model.blocks.{bid}.norm2z9vision_model.model.layers.{bid}.pre_feedforward_layernormzvision_tower.blocks.{bid}.norm2z-vision_model.transformer.resblocks.{bid}.ln_2zCmodel.qwen2_model.model.model.layers.{bid}.post_attention_layernorm)z<model.vision_tower.vision_model.encoder.layers.{bid}.mlp.fc1z6vision_tower.vision_model.encoder.layers.{bid}.mlp.fc1z/model.vision_tower.encoder.layers.{bid}.mlp.fc1z.model.vision_tower.encoder.layer.{bid}.mlp.fc1z vpm.encoder.layers.{bid}.mlp.fc1z/model.vision_model.encoder.layers.{bid}.mlp.fc1z"vit.layers.{bid}.mlp.dense_h_to_4hz:vision_tower.transformer.layers.{bid}.feed_forward.up_projz7vision_encoder.transformer.layers.{bid}.feed_forward.w3z'vision_model.model.layers.{bid}.mlp.fc1zvisual.blocks.{bid}.mlp.fc1zvisual.blocks.{bid}.mlp.up_projz"visual.blocks.{bid}.mlp.linear_fc1z)vision_tower.encoder.blocks.{bid}.mlp.fc0z3model.vision_model.transformer.layers.{bid}.mlp.fc1z-model.vision.transformer.layers.{bid}.mlp.fc1z1siglip2.vision_model.encoder.layers.{bid}.mlp.fc1z3vision_model.radio_model.model.blocks.{bid}.mlp.fc1z+vision_model.model.layers.{bid}.mlp.up_projz1vision_model.transformer.resblocks.{bid}.mlp.c_fcz6model.qwen2_model.model.model.layers.{bid}.mlp.up_proj)z<vision_tower.transformer.layers.{bid}.feed_forward.gate_projz7vision_encoder.transformer.layers.{bid}.feed_forward.w1z!visual.blocks.{bid}.mlp.gate_projz-vision_model.model.layers.{bid}.mlp.gate_projz8model.qwen2_model.model.model.layers.{bid}.mlp.gate_proj)z<model.vision_tower.vision_model.encoder.layers.{bid}.mlp.fc2z6vision_tower.vision_model.encoder.layers.{bid}.mlp.fc2z/model.vision_tower.encoder.layers.{bid}.mlp.fc2z.model.vision_tower.encoder.layer.{bid}.mlp.fc2z vpm.encoder.layers.{bid}.mlp.fc2z/model.vision_model.encoder.layers.{bid}.mlp.fc2z"vit.layers.{bid}.mlp.dense_4h_to_hz<vision_tower.transformer.layers.{bid}.feed_forward.down_projz7vision_encoder.transformer.layers.{bid}.feed_forward.w2z'vision_model.model.layers.{bid}.mlp.fc2zvisual.blocks.{bid}.mlp.fc2z!visual.blocks.{bid}.mlp.down_projz"visual.blocks.{bid}.mlp.linear_fc2z)vision_tower.encoder.blocks.{bid}.mlp.fc1z-model.vision.transformer.layers.{bid}.mlp.fc2z3model.vision_model.transformer.layers.{bid}.mlp.fc2z1siglip2.vision_model.encoder.layers.{bid}.mlp.fc2z3vision_model.radio_model.model.blocks.{bid}.mlp.fc2z8model.qwen2_model.model.model.layers.{bid}.mlp.down_projz-vision_model.model.layers.{bid}.mlp.down_projz3vision_model.transformer.resblocks.{bid}.mlp.c_proj)rC   )z:vision_model.model.layers.{bid}.post_feedforward_layernorm)z2vision_tower.vision_model.encoder.layers.{bid}.ls1z/model.vision_tower.encoder.layer.{bid}.lambda_1z-vision_model.transformer.resblocks.{bid}.ls_1)z2vision_tower.vision_model.encoder.layers.{bid}.ls2z/model.vision_tower.encoder.layer.{bid}.lambda_2z-vision_model.transformer.resblocks.{bid}.ls_2)z,vision_model.model.layers.{bid}.layer_scalar)z&vision_tower.vision_model.pre_layrnormzvision_tower.ln_prezvision_encoder.ln_prezvision_model.layernorm_prezmodel.vision_model.pre_layrnormrB   zvision_model.ln_pre)
z.model.vision_tower.vision_model.post_layernormz(vision_tower.vision_model.post_layernormz!model.vision_tower.post_layernormz!model.vision_model.post_layernormzvision_model.layernorm_postzvisual.merger.ln_qz$vision_tower.encoder.final_layernormzvisual.post_layernormz#siglip2.vision_model.post_layernormz"model.qwen2_model.model.model.norm)z"visual.merger.post_projection_normzvision_tower.post_trunk_normzvit.perceive.after_rms)z)multi_modal_projector.mm_input_projection)
zmulti_modal_projector.normz multi_modal_projector.layer_normzmulti_modal_projector.pre_normzmm_projector.pre_normpre_mm_projector_normmodel.vision.linear_proj.norm1zmlp_AR.pre_normzmerger.ln_qzvision_tower.merger.ln_qzmodel.merger.mlp.0.pre_norm)z&multi_modal_projector.mm_soft_emb_norm)zresampler.pos_embed_k)zresampler.attn.in_proj_q)zresampler.attn.in_proj_k)zresampler.attn.in_proj_v)zresampler.attn.out_projz*model.vision_model.head.attention.out_proj)zresampler.kv_proj)zresampler.ln_post)zresampler.ln_kv)zresampler.ln_q)zresampler.proj)zresampler.query)zv.token_embd.img_break)z0multi_modal_projector.patch_merger.merging_layerzpatch_merger.merging_layerzvisual.downsample)z-model.visual.deepstack_merger_list.{bid}.norm)z3model.visual.deepstack_merger_list.{bid}.linear_fc1)z3model.visual.deepstack_merger_list.{bid}.linear_fc2)z)model.vision_tower.vit_merger.layer_norm1)z.model.vision_tower.vit_merger.self_attn.q_proj)z.model.vision_tower.vit_merger.self_attn.k_proj)z.model.vision_tower.vit_merger.self_attn.v_proj)z0model.vision_tower.vit_merger.self_attn.out_proj)z&model.vision_tower.vit_merger.pre_norm)z&model.vision_tower.vit_merger.linear_1)z&model.vision_tower.vit_merger.linear_2)zmodel.sam_model.pos_embed)z model.sam_model.patch_embed.proj)z"model.sam_model.blocks.{bid}.norm1)z"model.sam_model.blocks.{bid}.norm2)z+model.sam_model.blocks.{bid}.attn.rel_pos_h)z+model.sam_model.blocks.{bid}.attn.rel_pos_w)z%model.sam_model.blocks.{bid}.attn.qkv)z&model.sam_model.blocks.{bid}.attn.proj)z%model.sam_model.blocks.{bid}.mlp.lin1)z%model.sam_model.blocks.{bid}.mlp.lin2)zmodel.sam_model.neck.{bid})zmodel.sam_model.net_2)zmodel.sam_model.net_3)zmodel.qwen2_model.query_768)zmodel.qwen2_model.query_1024)rE   )z&model.vision.linear_proj.dense_h_to_4hzvisual.merger.up_projzmodel.merger.mlp.0.linear_1)z&model.vision.linear_proj.dense_4h_to_hzvisual.merger.down_projzmodel.merger.mlp.0.linear_2)z"model.vision.linear_proj.gate_projzvisual.merger.gate_proj)zmodel.vision.boi)zmodel.vision.eoi)zvit.perceive.before_rms)zvit.perceive.image_begin)zvit.perceive.image_end)zmodel.vision_tower.std_bias)zmodel.vision_tower.std_scale)z0model.layerwise_projectors.{bid}.image_positionsz.model.spatial_projectors.{bid}.image_positions)z&model.layerwise_projectors.{bid}.queryz$model.spatial_projectors.{bid}.query)z+model.layerwise_projectors.{bid}.out_linearz)model.spatial_projectors.{bid}.out_linear)z%model.layerwise_projectors.{bid}.normz#model.spatial_projectors.{bid}.norm)z2model.layerwise_projectors.{bid}.qformer.layernormz0model.spatial_projectors.{bid}.qformer.layernorm)zPmodel.layerwise_projectors.qformer.encoder.layer.{bid}.attention.attention.queryzNmodel.spatial_projectors.qformer.encoder.layer.{bid}.attention.attention.query)zNmodel.layerwise_projectors.qformer.encoder.layer.{bid}.attention.attention.keyzLmodel.spatial_projectors.qformer.encoder.layer.{bid}.attention.attention.key)zPmodel.layerwise_projectors.qformer.encoder.layer.{bid}.attention.attention.valuezNmodel.spatial_projectors.qformer.encoder.layer.{bid}.attention.attention.value)zMmodel.layerwise_projectors.qformer.encoder.layer.{bid}.attention.output.densezKmodel.spatial_projectors.qformer.encoder.layer.{bid}.attention.output.dense)zQmodel.layerwise_projectors.qformer.encoder.layer.{bid}.attention.output.LayerNormzOmodel.spatial_projectors.qformer.encoder.layer.{bid}.attention.output.LayerNorm)zUmodel.layerwise_projectors.qformer.encoder.layer.{bid}.crossattention.attention.queryzSmodel.spatial_projectors.qformer.encoder.layer.{bid}.crossattention.attention.query)zSmodel.layerwise_projectors.qformer.encoder.layer.{bid}.crossattention.attention.keyzQmodel.spatial_projectors.qformer.encoder.layer.{bid}.crossattention.attention.key)zUmodel.layerwise_projectors.qformer.encoder.layer.{bid}.crossattention.attention.valuezSmodel.spatial_projectors.qformer.encoder.layer.{bid}.crossattention.attention.value)zRmodel.layerwise_projectors.qformer.encoder.layer.{bid}.crossattention.output.densezPmodel.spatial_projectors.qformer.encoder.layer.{bid}.crossattention.output.dense)zVmodel.layerwise_projectors.qformer.encoder.layer.{bid}.crossattention.output.LayerNormzTmodel.spatial_projectors.qformer.encoder.layer.{bid}.crossattention.output.LayerNorm)zOmodel.layerwise_projectors.qformer.encoder.layer.{bid}.intermediate_query.densezMmodel.spatial_projectors.qformer.encoder.layer.{bid}.intermediate_query.dense)zImodel.layerwise_projectors.qformer.encoder.layer.{bid}.output_query.densezGmodel.spatial_projectors.qformer.encoder.layer.{bid}.output_query.dense)zMmodel.layerwise_projectors.qformer.encoder.layer.{bid}.output_query.LayerNormzKmodel.spatial_projectors.qformer.encoder.layer.{bid}.output_query.LayerNorm)zaudio_tower.embed_positionszaudio_embedding.embedding)zaudio_embedding.embedding_norm)zaudio_embedding.to_logits)zaudio_tower.conv{bid}zconformer.pre_encode.conv.{bid}z;model.audio_tower.subsample_conv_projection.conv_{bid}.convz3conformer.subsample_conv_projection.layer{bid}.conv)z3conformer.subsample_conv_projection.layer{bid}.norm)z5conformer.subsample_conv_projection.input_proj_linearzencoder.input_linear)zaudio_tower.conv2d{bid})zaudio_tower.conv_out)zaudio_tower.layer_normzaudio_tower.ln_post)z)audio_tower.layers.{bid}.self_attn.q_projz)conformer.layers.{bid}.self_attn.linear_qz,conformer.layers.{bid}.attention.attn.q_projz'conformer.layers.{bid}.self_attn.q_projzencoder.layers.{bid}.attn.to_q)z)audio_tower.layers.{bid}.self_attn.k_projz)conformer.layers.{bid}.self_attn.linear_kz,conformer.layers.{bid}.attention.attn.k_projz'conformer.layers.{bid}.self_attn.k_projzencoder.layers.{bid}.attn.to_k)z)audio_tower.layers.{bid}.self_attn.v_projz)conformer.layers.{bid}.self_attn.linear_vz,conformer.layers.{bid}.attention.attn.v_projz'conformer.layers.{bid}.self_attn.v_projzencoder.layers.{bid}.attn.to_v)z0conformer.layers.{bid}.self_attn.relative_k_proj)z%conformer.layers.{bid}.norm_post_attn)z$conformer.layers.{bid}.norm_pre_attn)3conformer.layers.{bid}.attention.attn.per_dim_scalez.conformer.layers.{bid}.self_attn.per_dim_scale)zconformer.layers.{bid}.norm)z-audio_tower.layers.{bid}.self_attn_layer_normz$conformer.layers.{bid}.norm_self_attz.conformer.layers.{bid}.attention.pre_attn_normz"encoder.layers.{bid}.attn.pre_norm)z+audio_tower.layers.{bid}.self_attn.out_projz+conformer.layers.{bid}.self_attn.linear_outz%conformer.layers.{bid}.attention.postz%conformer.layers.{bid}.self_attn.postz encoder.layers.{bid}.attn.to_out)z)audio_tower.layers.{bid}.final_layer_normzconformer.layers.{bid}.norm_outz*conformer.layers.{bid}.attention.post_normzencoder.layers.{bid}.post_norm)z)conformer.layers.{bid}.norm_feed_forward1z5conformer.layers.{bid}.ffw_layer_start.pre_layer_normz3conformer.layers.{bid}.feed_forward1.pre_layer_normz!encoder.layers.{bid}.ff1.pre_norm)z6conformer.layers.{bid}.ffw_layer_start.post_layer_normz4conformer.layers.{bid}.feed_forward1.post_layer_norm)z7conformer.layers.{bid}.ffw_layer_start.post_layer_scale)zaudio_tower.layers.{bid}.fc1z,conformer.layers.{bid}.feed_forward1.linear1z2conformer.layers.{bid}.ffw_layer_start.ffw_layer_1z0conformer.layers.{bid}.feed_forward1.ffw_layer_1z encoder.layers.{bid}.ff1.up_proj)zaudio_tower.layers.{bid}.fc2z,conformer.layers.{bid}.feed_forward1.linear2z2conformer.layers.{bid}.ffw_layer_start.ffw_layer_2z0conformer.layers.{bid}.feed_forward1.ffw_layer_2z"encoder.layers.{bid}.ff1.down_proj)z,conformer.layers.{bid}.feed_forward2.linear1z0conformer.layers.{bid}.ffw_layer_end.ffw_layer_1z0conformer.layers.{bid}.feed_forward2.ffw_layer_1z encoder.layers.{bid}.ff2.up_proj)z,conformer.layers.{bid}.feed_forward2.linear2z0conformer.layers.{bid}.ffw_layer_end.ffw_layer_2z0conformer.layers.{bid}.feed_forward2.ffw_layer_2z"encoder.layers.{bid}.ff2.down_proj)z)conformer.layers.{bid}.norm_feed_forward2z3conformer.layers.{bid}.ffw_layer_end.pre_layer_normz3conformer.layers.{bid}.feed_forward2.pre_layer_normz!encoder.layers.{bid}.ff2.pre_norm)z4conformer.layers.{bid}.ffw_layer_end.post_layer_normz4conformer.layers.{bid}.feed_forward2.post_layer_norm)z5conformer.layers.{bid}.ffw_layer_end.post_layer_scale)z+conformer.layers.{bid}.self_attn.linear_poszJconformer.layers.{bid}.attention.attn.relative_position_embedding.pos_proj)z+conformer.layers.{bid}.self_attn.pos_bias_u)z+conformer.layers.{bid}.self_attn.pos_bias_v)zconformer.pre_encode.outz=model.audio_tower.subsample_conv_projection.input_proj_linearzconformer.output_proj)z(audio.multi_modal_projector.linear_{bid}zaudio_adapter.model.{bid}zaudio_tower.proj{bid})z"audio.multi_modal_projector.linearzaudio_tower.projzmodel.audio_tower.output_proj)z"audio.multi_modal_projector.ln_pre)z"audio.multi_modal_projector.ln_mid)z*conformer.layers.{bid}.conv.depthwise_convz/conformer.layers.{bid}.lconv1d.depthwise_conv1dz)encoder.layers.{bid}.conv.depth_conv.conv)z&conformer.layers.{bid}.conv.batch_normz-conformer.layers.{bid}.lconv1d.pre_layer_normz$encoder.layers.{bid}.conv.batch_norm)z+conformer.layers.{bid}.conv.pointwise_conv1z+conformer.layers.{bid}.lconv1d.linear_startz!encoder.layers.{bid}.conv.up_conv)z+conformer.layers.{bid}.conv.pointwise_conv2z)conformer.layers.{bid}.lconv1d.linear_endz#encoder.layers.{bid}.conv.down_conv)z conformer.layers.{bid}.norm_convz(conformer.layers.{bid}.lconv1d.conv_normzencoder.layers.{bid}.conv.norm)z7conformer.layers.{bid}.attention.attn.per_dim_key_scale)rF   )zmodel.embed_audio.embedding)z%model.embed_audio.hard_embedding_norm)z&model.embed_audio.embedding_projection)z%model.embed_audio.soft_embedding_norm)z,encoder.layers.{bid}.attn.rel_pos_emb.weight)z?projector.qformer.encoder.layer.{bid}.attention.attention.query)z=projector.qformer.encoder.layer.{bid}.attention.attention.key)z?projector.qformer.encoder.layer.{bid}.attention.attention.value)z<projector.qformer.encoder.layer.{bid}.attention.output.dense)z@projector.qformer.encoder.layer.{bid}.attention.output.LayerNorm)zDprojector.qformer.encoder.layer.{bid}.crossattention.attention.query)zBprojector.qformer.encoder.layer.{bid}.crossattention.attention.key)zDprojector.qformer.encoder.layer.{bid}.crossattention.attention.value)zAprojector.qformer.encoder.layer.{bid}.crossattention.output.dense)zEprojector.qformer.encoder.layer.{bid}.crossattention.output.LayerNorm)z>projector.qformer.encoder.layer.{bid}.intermediate_query.dense)z8projector.qformer.encoder.layer.{bid}.output_query.dense)z<projector.qformer.encoder.layer.{bid}.output_query.LayerNorm)pre_projection)post_projection)zmodel.layers.{bid}.eh_proj)zmodel.layers.{bid}.embed_tokens)zmodel.layers.{bid}.enorm)zmodel.layers.{bid}.hnorm)z#model.layers.{bid}.shared_head.head)z#model.layers.{bid}.shared_head.normblock_mappings_cfg)z%model.layers.{bid}.residual_layernorm)r)   z5dict[MODEL_ARCH, dict[MODEL_TENSOR, tuple[str, ...]]]arch_block_mappings_cfgz#dict[str, tuple[MODEL_TENSOR, str]]mappingarchr   n_blocksintc                L   i | _         | j                                        D ]?\  }}|t          |         vrt          |         }||f| j         |<   |D ]}||f| j         |<   @|| j        v r%| j                            | j        |                    t          |          D ]}| j                                        D ]i\  }}|t          |         vrt          |         	                    |          }||f| j         |<   |D ]$}|	                    |          }||f| j         |<   %jd S )N)bid)
rK   r$   itemsr   r	   rJ   rI   updaterangeformat)selfrL   rM   tensorkeystensor_namekeyrP   s           A/home/rurouni/src/llama.cpp-master/gguf-py/gguf/tensor_mapping.py__init__zTensorNameMap.__init__y	  sn    -3355 	: 	:LFD]4000&v.K)/(=DL% : :%+[$9S!!:4///#**4+G+MNNN?? 		> 		>C $ 7 = = ? ? > >t!444*62999DD-3[,A[) > >C**3*//C)/(=DL%%>>		> 		>    rY   strtry_suffixesSequence[str]returntuple[MODEL_TENSOR, str] | Nonec                   | j                             |          }||S |D ]^}|                    |          rG| j                             |d t          |                              }||d         |d         |z   fc S _d S )Nr   r   )rK   getendswithlen)rU   rY   r^   resultsuffixs        rZ   get_type_and_namezTensorNameMap.get_type_and_name	  s    !!#&&M" 	9 	9F||F## 9))#mF|m*<==%!!9fQi&&88888tr\   
str | Nonec                H    |                      ||          }|d S |d         S )Nr^   r   rh   rU   rY   r^   rf   s       rZ   get_namezTensorNameMap.get_name	  .    ''L'II>4ayr\   MODEL_TENSOR | Nonec                H    |                      ||          }|d S |d         S )Nrk   r   rl   rm   s       rZ   get_typezTensorNameMap.get_type	  ro   r\   c                d    	 | j         |         d         S # t          $ r t          |          w xY w)Nr   )rK   KeyErrorrU   rY   s     rZ   __getitem__zTensorNameMap.__getitem__	  s@    	 <$Q'' 	  	  	 3--	 s    /boolc                    || j         v S N)rK   ru   s     rZ   __contains__zTensorNameMap.__contains__	  s    dl""r\   c                *    t          | j                  S ry   )reprrK   )rU   s    rZ   __repr__zTensorNameMap.__repr__	  s    DL!!!r\   N)rL   r   rM   rN   )r#   )rY   r]   r^   r_   r`   ra   )rY   r]   r^   r_   r`   ri   )rY   r]   r^   r_   r`   rp   )rY   r]   r`   r]   )rY   r]   r`   rw   )r`   r]   (  __name__
__module____qualname__r   
TOKEN_EMBDMASKED_EMBD_CENTROIDSMASKED_EMBD_ORDERINGTOKEN_TYPESTOKEN_EMBD_NORMPOS_EMBDOUTPUTDENSE_2_OUTDENSE_3_OUTOUTPUT_NORM
ROPE_FREQSROPE_FACTORS_LONGROPE_FACTORS_SHORTCONV1DV_MM_EMBEDDINGV_MM_HARD_EMB_NORMV_MM_INP_PROJV_MM_SOFT_EMB_NORMV_ENC_CONV_STEMV_ENC_CONV_STEM_NORMV_ENC_MSFA_EXPV_ENC_MSFA_EXP_NORMV_ENC_MSFA_PROJV_ENC_MSFA_PROJ_NORMV_ENC_MSFA_NORM	A_CTC_OUTA_CTC_OUT_MIDA_QF_PROJ_QUERYA_QF_PROJ_NORMA_QF_PROJ_LINEARr$   __annotations__	ATTN_NORMATTN_NORM_2ATTN_QKVATTN_QATTN_KATTN_VATTN_OUTATTN_OUT_NORMATTN_POST_NORMATTN_ROT_EMBD
ATTN_SINKS	ATTN_GATEFFN_NORMFFN_PRE_NORMFFN_PRE_NORM_2FFN_POST_NORMFFN_POST_NORM_1FFN_POST_NORM_2FFN_GATE_INPFFN_GATE_INP_SHEXPFFN_EXP_PROBS_BFFN_UP
FFN_UP_EXPFFN_UP_SHEXPFFN_UP_CHEXPFFN_ACTFFN_GATEFFN_GATE_EXPFFN_GATE_SHEXPFFN_GATE_CHEXPFFN_GATE_UP_EXPMOE_LATENT_DOWNMOE_LATENT_UPFFN_DOWNFFN_DOWN_EXPFFN_DOWN_SHEXPFFN_DOWN_CHEXPATTN_Q_NORMATTN_K_NORMLAYER_OUT_NORMLAYER_OUT_SCALEPER_LAYER_TOKEN_EMBDPER_LAYER_MODEL_PROJPER_LAYER_PROJ_NORM
ALTUP_PROJALTUP_UNEMBD_PROJPER_LAYER_INP_GATEPER_LAYER_PROJPER_LAYER_POST_NORMALTUP_CORRECT_COEFALTUP_CORRECT_SCALEALTUP_PREDICT_COEFALTUP_ROUTERALTUP_ROUTER_NORMLAUREL_LLAUREL_RLAUREL_POST_NORMSSM_IN
SSM_CONV1DSSM_XSSM_DTSSM_DT_NORMSSM_A
SSM_B_NORM
SSM_C_NORMSSM_DSSM_NORMSSM_OUT	SSM_ALPHASSM_BETA_ALPHASSM_CONV1D_QSSM_CONV1D_KSSM_CONV1D_VSSM_F_ASSM_F_BSSM_BETASSM_G_ASSM_G_BTIME_MIX_W0TIME_MIX_W1TIME_MIX_W2TIME_MIX_A0TIME_MIX_A1TIME_MIX_A2TIME_MIX_V0TIME_MIX_V1TIME_MIX_V2TIME_MIX_G1TIME_MIX_G2TIME_MIX_K_KTIME_MIX_K_ATIME_MIX_R_KTIME_MIX_LERP_XTIME_MIX_LERP_KTIME_MIX_LERP_VTIME_MIX_LERP_RTIME_MIX_LERP_GTIME_MIX_LERP_WTIME_MIX_FIRSTTIME_MIX_DECAYTIME_MIX_DECAY_W1TIME_MIX_DECAY_W2TIME_MIX_KEYTIME_MIX_VALUETIME_MIX_RECEPTANCETIME_MIX_GATETIME_MIX_LNTIME_MIX_OUTPUTCHANNEL_MIX_LERP_KCHANNEL_MIX_LERP_RCHANNEL_MIX_KEYCHANNEL_MIX_RECEPTANCECHANNEL_MIX_VALUEATTN_Q_AATTN_Q_BATTN_KV_A_MQA	ATTN_KV_BATTN_K_BATTN_V_BATTN_Q_A_NORMATTN_KV_A_NORMATTN_SUB_NORMFFN_SUB_NORMDEC_ATTN_NORM
DEC_ATTN_Q
DEC_ATTN_K
DEC_ATTN_VDEC_ATTN_OUTDEC_ATTN_REL_BDEC_CROSS_ATTN_NORMDEC_CROSS_ATTN_QDEC_CROSS_ATTN_KDEC_CROSS_ATTN_VDEC_CROSS_ATTN_OUTDEC_CROSS_ATTN_REL_BDEC_FFN_NORMDEC_FFN_GATE
DEC_FFN_UPDEC_FFN_DOWNDEC_OUTPUT_NORMENC_ATTN_NORM
ENC_ATTN_Q
ENC_ATTN_K
ENC_ATTN_VENC_ATTN_OUTENC_ATTN_REL_BENC_FFN_NORMENC_FFN_GATE
ENC_FFN_UPENC_FFN_DOWN	VISEXP_UPVISEXP_GATEVISEXP_DOWNVISEXP_ATTN_OUTVISEXP_ATTN_QKVINDEXER_K_NORMINDEXER_PROJINDEXER_ATTN_KINDEXER_ATTN_Q_BENC_OUTPUT_NORMCLSCLS_OUTCLS_NORMCONVNEXT_DWCONVNEXT_NORMCONVNEXT_PW1CONVNEXT_PW2CONVNEXT_GAMMAPOSNET_CONV1POSNET_CONV2POSNET_NORMPOSNET_NORM1POSNET_NORM2POSNET_ATTN_NORMPOSNET_ATTN_QPOSNET_ATTN_KPOSNET_ATTN_VPOSNET_ATTN_OUTSHORTCONV_CONVSHORTCONV_INPROJSHORTCONV_OUTPROJV_MMPROJV_MMPROJ_FCV_MMPROJ_MLPV_MMPROJ_PEGV_ENC_EMBD_CLSV_ENC_EMBD_PATCHV_ENC_EMBD_NORMV_ENC_EMBD_PATCH_NORMV_ENC_EMBD_POSV_ENC_EMBD_IMGNLV_ENC_EMBD_VSEPV_ENC_ATTN_QKVV_ENC_ATTN_QV_ENC_ATTN_Q_NORMV_ENC_ATTN_KV_ENC_ATTN_K_NORMV_ENC_ATTN_VV_ENC_INPUT_NORMV_ENC_ATTN_OV_ENC_ATTN_SINKSV_ENC_POST_ATTN_NORMV_ENC_FFN_UPV_ENC_FFN_GATEV_ENC_FFN_DOWNV_ENC_ATTN_POST_NORMV_ENC_FFN_POST_NORMV_LAYER_SCALE_1V_LAYER_SCALE_2V_LAYER_OUT_SCALE
V_PRE_NORMV_POST_NORMV_MM_POST_NORMV_MM_INP_NORMV_RESMPL_POS_EMBD_KV_RESMPL_ATTN_QV_RESMPL_ATTN_KV_RESMPL_ATTN_VV_RESMPL_ATTN_OUTV_RESMPL_KVV_RESMPL_POST_NORMV_RESMPL_KV_NORMV_RESMPL_Q_NORMV_RESMPL_PROJV_RESMPL_QUERYV_TOK_EMBD_IMG_BREAKV_MM_PATCH_MERGER	V_DS_NORMV_DS_FC1V_DS_FC2V_MERGER_LN1V_MERGER_ATTN_QV_MERGER_ATTN_KV_MERGER_ATTN_VV_MERGER_ATTN_OV_MERGER_DS_LNV_MERGER_DS_UPV_MERGER_DS_DOWNV_SAM_POS_EMBDV_SAM_PATCH_EMBDV_SAM_PRE_NORMV_SAM_POST_NORMV_SAM_ATTN_POS_HV_SAM_ATTN_POS_WV_SAM_ATTN_QKVV_SAM_ATTN_OUTV_SAM_MLP_LIN_1V_SAM_MLP_LIN_2
V_SAM_NECKV_SAM_NET_2V_SAM_NET_3V_RESMPL_QUERY_768V_RESMPL_QUERY_1024V_MM_POST_FC_NORMV_MM_UP	V_MM_DOWN	V_MM_GATE	V_TOK_BOI	V_TOK_EOIV_MM_PRE_NORMV_TOK_IMG_BEGINV_TOK_IMG_END
V_STD_BIASV_STD_SCALEV_MULTI_PROJ_IMG_POSV_MULTI_PROJ_QUERYV_MULTI_PROJ_LINEARV_MULTI_PROJ_NORMV_MULTI_PROJ_POST_NORMV_QF_SELF_ATTN_QV_QF_SELF_ATTN_KV_QF_SELF_ATTN_VV_QF_SELF_ATTN_OV_QF_SELF_ATTN_NORMV_QF_CROSS_ATTN_QV_QF_CROSS_ATTN_KV_QF_CROSS_ATTN_VV_QF_CROSS_ATTN_OV_QF_CROSS_ATTN_NORMV_QF_FFN_UPV_QF_FFN_DOWNV_QF_FFN_NORMA_ENC_EMBD_POSA_ENC_EMBD_NORMA_ENC_EMBD_TO_LOGITSA_ENC_CONV1DA_ENC_CONV1D_NORMA_ENC_INP_PROJA_ENC_CONV2DA_ENC_CONV_OUT
A_PRE_NORMA_POST_NORMA_ENC_ATTN_QA_ENC_ATTN_KA_ENC_ATTN_VA_ENC_ATTN_K_RELA_ENC_ATTN_POST_NORMA_ENC_ATTN_PRE_NORMA_ENC_PER_DIM_SCALEA_ENC_LAYER_PRE_NORMA_ENC_INPUT_NORMA_ENC_OUTPUTA_ENC_OUTPUT_NORMA_ENC_FFN_NORMA_ENC_FFN_POST_NORMA_ENC_FFN_SCALEA_ENC_FFN_UPA_ENC_FFN_GATEA_ENC_FFN_DOWNA_ENC_FFN_UP_1A_ENC_FFN_DOWN_1A_ENC_FFN_NORM_1A_ENC_FFN_POST_NORM_1A_ENC_FFN_SCALE_1A_ENC_LINEAR_POSA_ENC_POS_BIAS_UA_ENC_POS_BIAS_V	A_ENC_OUTA_MMPROJA_MMPROJ_FCA_MM_NORM_PREA_MM_NORM_MIDA_ENC_CONV_DWA_ENC_CONV_NORMA_ENC_CONV_PW1A_ENC_CONV_PW2A_ENC_NORM_CONVA_PER_DIM_K_SCALEA_PER_DIM_SCALEA_MM_EMBEDDINGA_MM_HARD_EMB_NORMA_MM_INP_PROJA_MM_SOFT_EMB_NORMA_ENC_ATTN_REL_POS_EMBA_QF_SELF_ATTN_QA_QF_SELF_ATTN_KA_QF_SELF_ATTN_VA_QF_SELF_ATTN_OA_QF_SELF_ATTN_NORMA_QF_CROSS_ATTN_QA_QF_CROSS_ATTN_KA_QF_CROSS_ATTN_VA_QF_CROSS_ATTN_OA_QF_CROSS_ATTN_NORMA_QF_FFN_UPA_QF_FFN_DOWNA_QF_FFN_NORMNEXTN_PROJ_PRENEXTN_PROJ_POSTNEXTN_EH_PROJNEXTN_EMBED_TOKENSNEXTN_ENORMNEXTN_HNORMNEXTN_SHARED_HEAD_HEADNEXTN_SHARED_HEAD_NORMrI   r   ARCTICFFN_NORM_EXPrJ   r[   rh   rn   rr   rv   rz   r}   r#   r\   rZ   r   r      s!        m9 "
m9@ 	* -
Am9F 	) ,
Gm9P 	  #
Qm9Z 	$ '
[m9x 	  
ym9H 	 
Im9b 	  #
cm9h 	  #
im9p 	  #
qm9b 	 "
cm9l 	&mm9n 	'om9r 	 
sm9z 	# &
{m9@ 	' *
Am9F 	" %
Gm9L 	' *
 	$ '
 	) ,
 	# &
 	( +
 	$ '
 	) ,
 	$ '
 	 !
 	" %
 	$ '
 	# &
 	% (
Um9 m9L m m m m^q"? !!
q"?L 	  #
Mq"?^ 	  
_q"?N 	 
Oq"?v 	 
wq"?` 	 
aq"?H 	 $ 
Iq"?V 	" %
Wq"?h 	# &
iq"?x 	" %
yq"?F 	 "
Gq"?P 	 !
Qq"?^ 	  
_q"?T 	! $
Uq"?b 	# &
cq"?l 	" %
mq"?~ 	$ '
q"? q"?F 	$ '
Gq"?N 	! $
Oq"?v 	' *
wq"?~ 	$ '
q"?Z	 	 (
[	q"?n
 	 
"
o
q"?F 	! 
$
Gq"?^ 	! $
_q"?h 	 
iq"?r 	  
sq"?Z 	! 	$
[q"?p 	# &
qq"?D 	# &
Eq"?L 	$ '
Mq"?V 	$ '
Wq"?^ 	" %
_q"?h 	 # 
iq"? q"? q"?r 	! $
sq"?N 	# 
&
Oq"?f 	# &
gq"?n 	  #
oq"?P 	  #
Qq"?p 	 "
qq"?x 	# &
yq"?L 	$ '
Mq"?V 	) ,
Wq"?^ 	) ,
_q"?f 	( +
gq"?n 	 "
oq"?v 	& )
wq"?~ 	' *
q"?F 	# &
Gq"?N 	( +
Oq"?V 	' *
Wq"? q"? q"?^ 	( +
_q"?f 	' *
gq"?n 	! $
oq"?v 	& )
wq"?~ 	  
q"?F 	  
Gq"?N 	% (
Oq"?V 	 
Wq"?f 	 "
gq"?v 	 
wq"?D 	 
Eq"?X 	  #
Yq"?b 	 
cq"?t 	 "
uq"?@ 	 "
Aq"?L 	 
Mq"?Z 	  
[q"? q"? q"?h 	 
iq"?x 	 !
yq"?@ 	# &
Aq"?J 	! $
Kq"?P 	! $
Qq"?V 	! $
Wq"?\ 	 
]q"?b 	 
cq"?h 	  
iq"?p 	 
qq"?v 	 
wq"?| 	  #
}q"?D 	  #
Eq"?P 	  #
Qq"?\ 	  #
]q"?d 	  #
eq"?l 	  #
mq"? q"? q"?t 	  #
uq"?| 	  #
}q"?D 	  #
Eq"?L 	  #
Mq"?T 	  #
Uq"?\ 	! $
]q"?d 	! $
eq"?l 	! $
mq"?t 	$ '
uq"?~ 	$ '
q"?H 	$ '
Iq"?R 	$ '
Sq"?\ 	$ '
]q"?f 	$ '
gq"?p 	# &
qq"?x 	# &
yq"?B 	& )
Cq"? q"? q"?L 	& )
Mq"?V 	! $
Wq"?d 	# &
eq"?r 	( +
sq"?@ 	" %
Aq"?J 	  #
Kq"?T 	$ '
Uq"?b 	' *
cq"?l 	' *
mq"?t 	$ '
uq"?~ 	+ .
q"?F 	& )
Gq"?P 	  
Qq"?Z 	  
[q"?d 	" %
eq"?n 	 !
oq"?v 	  
wq"? q"? q"?@ 	  
Aq"?J 	" %
Kq"?T 	# &
Uq"?^ 	" %
_q"?f 	! $
gq"?n 	" %
oq"?v 	 "
wq"?~ 	 "
q"?F 	 "
Gq"?N 	! $
Oq"?V 	# &
Wq"?^ 	( +
_q"?f 	% (
gq"?n 	% (
oq"?v 	% (
wq"?~ 	' *
q"?F 	) ,
Gq"? q"? q"?N 	! $
Oq"?V 	! $
Wq"?^ 	 "
_q"?h 	! $
iq"?p 	$ '
qq"?x 	" %
yq"?@  	 "
A q"?H  	 "
I q"?P  	 "
Q q"?X  	! $
Y q"?`  	# &
a q"?h  	! $
i q"?p  	! $
q q"?x  	 "
y q"?B! 	! $
C!q"?J! 	 !
K!q"?R! 	  #
S!q"? q"? q"?Z! 	  #
[!q"?b! 	$ '
c!q"?j! 	$ '
k!q"?r! 	# &
s!q"?z! 	! $
{!q"?B" 	# &
C"q"?J" 	% (
K"q"?V" 	$ '
W"q"?`" 	 
a"q"?p" 	 
q"q"?x" 	  
y"q"?B# 	  #
C#q"?J# 	" %
K#q"?R# 	! $
S#q"?Z# 	! $
[#q"?b# 	# &
c#q"?j# 	! $
k#q"? q"? q"?r# 	! $
s#q"?z# 	  #
{#q"?B$ 	! $
C$q"?J$ 	! $
K$q"?R$ 	% (
S$q"?Z$ 	" %
[$q"?b$ 	" %
c$q"?j$ 	" %
k$q"?r$ 	$ '
s$q"?z$ 	# &
{$q"?B% 	% (
C%q"?J% 	& )
K%q"?X% 	  
Y%q"?l% 	  #
m%q"?|% 	! $
}%q"?J& 	! $
K&q"?R& 	# &
S&q"? q"? q"?d& 	% (
e&q"?R' 	$ '
S'q"?\' 	* -
]'q"?d' 	# &
e'q"?L( 	% (
M(q"?V( 	$ '
W(q"?`( 	# &
a(q"?t( 	! $
u(q"?Z) 	& )
[)q"?h) 	! $
i)q"?N* 	& )
O*q"?\* 	! $
]*q"?B+ 	% (
C+q"?p+ 	! $
q+q"?b, 	% (
c,q"?j, 	) ,
k,q"?Z- 	! $
[-q"? q"? q"?J. 	# &
K.q"?Z. 	# &
[.q"?J/ 	) ,
K/q"?R/ 	( +
S/q"?Z/ 	$ '
[/q"?f/ 	$ '
g/q"?r/ 	& )
s/q"?z/ 	 "
{/q"?N0 	  #
O0q"?h0 	# &
i0q"?t0 	" %
u0q"?|0 	" %
}0q"?V1 	' *
W1q"?^1 	( +
_1q"?f1 	$ '
g1q"?n1 	$ '
o1q"?v1 	$ '
w1q"? q"? q"?~1 	& )
1q"?H2 	  #
I2q"?P2 	' *
Q2q"?X2 	% (
Y2q"?`2 	$ '
a2q"?h2 	" %
i2q"?p2 	# &
q2q"?x2 	) ,
y2q"?@3 	& )
A3q"?L3 	 !
M3q"?T3 	  
U3q"?\3 	  
]3q"?d3 	! $
e3q"?l3 	$ '
m3q"?t3 	$ '
u3q"?|3 	$ '
}3q"?D4 	$ '
E4q"? q"? q"?L4 	# &
M4q"?T4 	# &
U4q"?\4 	% (
]4q"?d4 	# &
e4q"?l4 	% (
m4q"?t4 	# &
u4q"?|4 	$ '
}4q"?D5 	% (
E5q"?L5 	% (
M5q"?T5 	# &
U5q"?\5 	# &
]5q"?d5 	$ '
e5q"?l5 	$ '
m5q"?t5 	 "
u5q"?|5 	  #
}5q"?D6 	  #
E6q"?L6 	' *
M6q"? q"? q"?T6 	( +
U6q"?\6 	& )
]6q"?d6 	 
e6q"?p6 	 !
q6q"?|6 	 !
}6q"?F7 	 !
G7q"?N7 	 !
O7q"?V7 	" %
W7q"?^7 	$ '
_7q"?f7 	" %
g7q"?n7 	 "
o7q"?v7 	  #
w7q"?@8 	) ,
A8q"?H8 	' *
I8q"?P8 	( +
Q8q"?X8 	& )
Y8q"?`8 	+ .
a8q"? q"? q"?l8 	% (
m8q"?t8 	% (
u8q"?|8 	% (
}8q"?D9 	% (
E9q"?L9 	( +
M9q"?T9 	& )
U9q"?\9 	& )
]9q"?d9 	& )
e9q"?l9 	& )
m9q"?t9 	) ,
u9q"?|9 	  #
}9q"?D: 	" %
E:q"?L: 	" %
M:q"?Z: 	# &
[:q"?d: 	$ '
e:q"?l: 	) ,
m:q"?t: 	! $
u:q"? q"? q"?B; 	& )
C;q"?J; 	# &
K;q"?T; 	! $
U;q"?\; 	# &
];q"?d; 	e;q"?h; 	  #
i;q"?r; 	! $
s;q"?B< 	! $
C<q"?R< 	! $
S<q"?b< 	% (
c<q"?j< 	) ,
k<q"?r< 	( +
s<q"?z< 	( +
{<q"?D= 	) ,
E=q"?L= 	% (
M=q"?Z= 	! $
[=q"?j= 	& )
k=q"? q"? q"?x= 	# &
y=q"?F> 	( +
G>q"?P> 	$ '
Q>q"?X> 	! $
Y>q"?h> 	#Ri>q"?l> 	# &
m>q"?|> 	# &
}>q"?J? 	% (
K?q"?X? 	% (
Y?q"?f? 	* -
g?q"?p? 	& )
q?q"?x? 	% (
y?q"?B@ 	% (
C@q"?J@ 	% (
K@q"?R@ 	 !
S@q"?d@ 	  
e@q"?p@ 	  #
q@q"? q"? q"?|@ 	" %
}@q"?DA 	" %
EAq"?LA 	" %
MAq"?XA 	$ '
YAq"?dA 	# &
eAq"?pA 	# &
qAq"?|A 	$ '
}Aq"?HB 	& )
IBq"?PB 	$ '
QBq"?XB 	# &
YBq"?^B 	' *
_Bq"?dB 	" %
eBq"?jB 	' *
kBq"?rB 	+ .
sBq"?zB 	% (
{Bq"?BC 	% (
CCq"?JC 	% (
KCq"? q"? q"?RC 	% (
SCq"?ZC 	( +
[Cq"?bC 	& )
cCq"?jC 	& )
kCq"?rC 	& )
sCq"?zC 	& )
{Cq"?BD 	) ,
CDq"?JD 	  #
KDq"?RD 	" %
SDq"?ZD 	" %
[Dq"?dD 	# &
eDq"?lD 	$ '
mDq"?tD 	" %
uDq"?|D 	' *
}Dq"?DE 	  #
EEq"?LE 	  #
MEq"?TE 	+ .
UEq"? q"?\E 	+ .
]Eq"? q"? q" q" q" q"jE 	! $ % (	
	V 	 	 	 	 1000> > > >,	 	 	 	 	               # # # #" " " " " "r\   r   rL   r   rM   rN   r`   c                "    t          | |          S ry   )r   )rL   rM   s     rZ   get_tensor_name_mapr  	  s    x(((r\   N)rL   r   rM   rN   r`   r   )
__future__r   typingr   	constantsr   r   r   r	   r   r  r#   r\   rZ   <module>r
     s    " " " " " "       L L L L L L L L L L L Lh&" h&" h&" h&" h&" h&" h&" h&"VM) ) ) ) ) )r\   