
    |-j@                    h   d dl mZ d dlmZ d dlmZ d dlmZ ddlm	Z	m
Z
mZmZmZmZmZmZmZ erddlmZ ddlmZ i d	d
dd
ddddddddddddddddddddddddddddddi ddddddddd dd!d"d#d"d$d"d%d&d'd(d)d(d*d(d+d(d,d(d-d(d.d(d/d(i d0d(d1d2d3d2d4d2d5d2d6d2d7d8d9d8d:d8d;d8d<d8d=d8d>d8d?d8d@dAdBdAdCdDi dEdDdFdDdGdDdHdDdIdDdJdDdKdDdLdDdMdDdNdDdOdDdPdDdQdRdSdTdUdVdWdVdXdVdVdVdVdVdYZdZ Zd[ad\ Z	 dsdtdfZdudjZ	 	 	 dvdwdrZd[S )x    )annotations)defaultdict)deepcopy)TYPE_CHECKING   )	ChunkConcatenate"ErnieFuseAndSplitTextVisionExpertsMergeModulelistPrefixChange	TransposeWeightConverterWeightRenamingWeightTransformPreTrainedModel)HfQuantizerminimaxmixtral
minimax_m2afmoe	qwen2_moedeepseek_v2deepseek_v3dots1ernie4_5_moeglm4_moeglm4_moe_liteglm_moe_dsa	glm4v_moelongcat_flash
solar_open	qwen3_moeqwen3_omni_moeqwen3_omni_moe_thinker
qwen3_nexthunyuan_v1_moe	flex_olmoolmoe
exaone_moecohere2_moe
rt_detr_v2rt_detrpp_doclayout_v2pp_doclayout_v3sam3_tracker_videosam3_trackeraltclip_vision_modelclip_vision_modelchinese_clip_vision_modelclipseg_vision_modelmetaclip_2_vision_modelmlcd_visionmlcdsiglip_vision_modelsiglip2_vision_modelxclip_vision_modelclipseg_text_modelclip_text_modelmetaclip_2_text_modelsiglip_text_modelsiglip2_text_modelxclip_text_model	paligemmallava
aya_visiongot_ocr2gemma3internvlvipllavamistral3pp_chart2tablellava_next_video
llava_nextllava_onevisionPaliGemmaModel
LlavaModelAyaVisionModelGotOcr2ModelGemma3ModelInternVLModelVipLlavaModelMistral3ModelPPChart2TableModelLlavaNextModelLlavaNextVideoModelLlavaOnevisionModel	FuyuModelMllamaModelMaskFormerDetrDecoder	DetrModel"Qwen2_5_VLForConditionalGenerationQwen2VLForConditionalGenerationASTModelViTModel	BeitModel	DeiTModel)
IJepaModelViTMAEModelViTMSNModel
VivitModelc            (     <:   i dt          dddgt          d          g          t          dg d	t          d          g          t          d
d          gdt          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          gdt          dd          gdt          dd          t          dd          t          dd          t          dd          t          dd          t          dd          gd t          d!d"          t          d#d$          t          d%d&          t          d'd(          t          d)d*          t          d+d,          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          gd-t          d.d/          t          d0d1          t          d!d"          t          d#d$          t          d%d&          t          d'd(          t          d2d3          t          d4d5          t          d6d7          g	d8t          d9d:          t          d;d<          t          d=d>          t          d?d          gd@t          dAdB          t          dCdD          t          dEdF          t          dGd:          t          dHd<          t          dId>          t          dJdK          t          dLdM          t          dd          t          dNd          t          dOd          t          dPdQ          t          dRdS          gdTt          dUdV          gdWt          dGd:          t          dHd<          t          dId>          t          dXdY          t          dd          t          dd          t          dd          gdZt          dd[          t          d+d\          gd]t          d^d          gd_g t          d`da          t          dbdc          t          ddde          t          dfdg          t          dhdi          t          djdk          t          dldm          t          dndo          t          dpdq          t          drds          t          dtdu          t          dvdw          t          dxdy          t          dzd{          t          d|d}          t          d~d          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          ddgdt          d          t	          d          g          t          ddt          d          g          dt          dd          gdt          dd          t          dd          t          dd          t          dd          gdt          dd          t          dd          t          dd          t          dd          t          dd          gdt          dìĦ          gi dt          dƬĦ          gdt          dd          gdt          dd          t          dd          t          ddʬ          t          dd̬          t          dd          gdt          dd          t          dd          t          ddϬ          gdt          dd          t          dd          t          ddҬ          t          dd          gdt          ddƬ          gdt          dd          t          ddج          t          ddڬ          gdt          ddݬ          t          dd߬          t          dd          gdt          dd߬          t          dd          gdt          dd          gdt          dd          gdt          d          gdt          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          gdt          dd          gd t          dd          gdt          dd          gdt          dd          t          d	d
          gi dt          dd          t          dd          gdt          dd          gdt          dd          t          dd          t          dd          gdt          dd          t          dd          t          dd          gdt          d d!          t          d"d#gd$t          d          t	          d          g          t          d%gd&t          d          g          gd't          d(d)gd*t          d          t	          d          g          t          d+d,t          d          g          gd-t          d*d*t          dd.d/0          g          t          d,d,t          dd.d/0          g          gd1t          d d!          t          d2d3          t          d"d#gd$t          d          t	          d          g          t          d%d&t          d          g          gd4t          d5d6gd7t          d          t	          d          g          t          d8d9t          d          g          gd:t          dÐd;          t          d<d=          t          d>d?          t          d@dA          t          dBdC          t          dDdE          t          dFdG          t          dHdI          t          dJdK          t          dLdM          t          dNdOt          ddP          g          t          dQdRt          ddP          g          t          dSgdTdUgt          d          g          t          dVgdWdXgt          ddY          g          t          dZd[gd\d]gt          ddY          g          gd^t          d_d`          t          dadb          gdct          ddde          t          dfdg          t          d_d`          t          dadb          gdht          dfdg          t          d_d`          t          dadb          t          didj          gdkt          dldm          t          dndo          t          dpdq          t          drds          t          dtdu          t          dvdw          t          dxdy          t          dzd{          t          d|d}          t          d~d          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dg dt          d          g          t          dg dt          d          g          gdt          dd          gdt          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          g
dt          ddde          t          dd          t          dd          t          d_d`          t          dadb          t          dd          t          dd          t          dd          t          dd          t          ddæ          t          dĐdŦ          t          dƐdǦ          t          dȐdɦ          t          dʐd˦          t          d̐dͦ          t          dΐdϦ          gt          ddde          t          dd          t          d_d`          t          dadb          gt          dfdg          t          d_dЦ          t          dadѦ          t          didj          gt          dҐd          t          dӐdԦ          t          dgdt          d          g          t          dgdt          d          g          gt          dِdgd7t          d          t	          d          g          t          dېd9t          d          g          gt          dܐdݦ          t          dސdߦ          t          dd          t          dd          t          dd          t          d
d          t          dd          t          dd          t          dgg dt          d          g          g	t          dސd߬          t          dܐdݬ          t          dd          t          dd          t          dd          t          dg dt          d          g          gt          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          d d          t          dd          t          dd          t          dd          t          dd	          t          d
d          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          dd          t          d d!          t          d"d#          t          d$d%          t          d&d'          gt          d(d)          t          d*d+          t          d,d-          t          d.d/          t          d0d1          t          d2d3          t          d4d5          t          d6d7          t          d8d9          t          d:d;          t          d<d=          t          d>d?          t          d@dA          t          dBdC          t          dDdE          t          dFgg dGt          d          g          t          dHgg dIt          d          g          gt          dJdK          t          dLdM          gdN	} | dxx         | dO                                         z  cc<   | dOxx         t          dPdQ          t          dRdS          gz  cc<   t          dTdU          t          dVdW          t          dXdY          t          dZd[          t          d\d]          t          d^d_          t          d`da          t          dbdc          t          ddde          t          dfdg          t          dhdi          t          djdk          t          dldm          t          dndo          t          dpdq          t          drds          g| dt<   | dt                                         | du<   | d'                                         | dv<   | dvxx         t          dSdw          gz  cc<   | d                                         t          dxdy          t          dzd{          gz   | d|<   | d                                         | d}<   | d}xx         t          ddQ          t          ddS          t          dd~          gz  cc<   | d                                         | d<   | dxx         t          dd          gz  cc<   | d'                                         | d<   | dxx         t          dd          gz  cc<   | d'                                         | d<   | dxx         t          ddQ          t          dd          t          dd          gz  cc<   | d                                         | d<   | dxx         | d'                                         z  cc<   | d'                                         | d<   | dxx         t          dd          t          dd          gz  cc<   t          
                                D ]'\  }}|| v r
| |                                         | |<   (| S (  Nhrm_textzmlp.gate_up_proj.weightzmlp.gate_proj.weightzmlp.up_proj.weightr   )dim)source_patternstarget_patterns
operationszattn.gqkv_proj.weight)zself_attn.gate_proj.weightself_attn.q_proj.weightself_attn.k_proj.weightself_attn.v_proj.weightz\.attn\.o_proj\.z.self_attn.o_proj.rk   rl   ra   zencoder\.layer\.zlayers.zattention.queryq_projzattention.keyk_projzattention.valuev_projzattention.output.densezattention.o_projzintermediate.densezmlp.fc1zoutput.densezmlp.fc2ViTMSNForImageClassificationz
^encoder\.zvit.encoder.ViTMAEForPreTrainingBeitBackbonez
^fpn1\.0\.zfpn.fpn1.conv_transpose1.z
^fpn1\.1\.zfpn.fpn1.normalization.z
^fpn1\.3\.zfpn.fpn1.conv_transpose2.z
^fpn2\.0\.z	fpn.fpn2.z^encoder\.layer\.zbeit.encoder.layer.z^embeddings\.zbeit.embeddings.BeitForSemanticSegmentationz (?<!psp_modules\.[0-9]\.1\.)bn\.znormalization.z((?<!psp_modules\.[0-9]\.1\.)conv\.weightzconvolution.weightzdecode_head.bottleneck.zdecode_head.psp_bottleneck.z0decode_head\.psp_modules\.(\d+)\.1\.conv\.weightz9decode_head.psp_modules.blocks.\1.conv.convolution.weightz(decode_head\.psp_modules\.(\d+)\.1\.bn\.z5decode_head.psp_modules.blocks.\1.conv.normalization.lw_detrzattention.attention.queryzattention.q_projzattention.attention.keyzattention.k_projzattention.attention.valuezattention.v_projzattention.outputSegformerModelzencoder.patch_embeddings.(\d+).zstages.\1.patch_embeddings.zencoder.block.(\d+).zstages.\1.blocks.zencoder.layer_norm.(\d+)zstages.\1.layer_normzattention.self.queryzattention.self.keyzattention.self.valuezattention.self.srz/attention.sequence_reduction.sequence_reductionzattention.self.layer_normz'attention.sequence_reduction.layer_normz
mlp.dense1z
mlp.dense2layer_norm_1layernorm_beforelayer_norm_2layernorm_after SegformerForSemanticSegmentationzdecode_head.linear_czdecode_head.linear_projectionsswinz+attention.self.relative_position_bias_tablez=attention.relative_position_bias.relative_position_bias_tableSwinBackbonezswin.encoder.zswin.embeddings.altclipzlayer\.deepseek_v4z^embed\.weight$zembed_tokens.weightz^head\.weight$zlm_head.weightz^norm\.weight$znorm.weightz^hc_head_fn$zhc_head.hc_fnz^hc_head_base$zhc_head.hc_basez^hc_head_scale$zhc_head.hc_scalez^layers\.(\d+)\.attn_norm\.zlayers.\1.input_layernorm.z^layers\.(\d+)\.ffn_norm\.z#layers.\1.post_attention_layernorm.z^layers\.(\d+)\.hc_attn_fn$zlayers.\1.attn_hc.fnz^layers\.(\d+)\.hc_attn_base$zlayers.\1.attn_hc.basez^layers\.(\d+)\.hc_attn_scale$zlayers.\1.attn_hc.scalez^layers\.(\d+)\.hc_ffn_fn$zlayers.\1.ffn_hc.fnz^layers\.(\d+)\.hc_ffn_base$zlayers.\1.ffn_hc.basez^layers\.(\d+)\.hc_ffn_scale$zlayers.\1.ffn_hc.scalez^layers\.(\d+)\.attn\.zlayers.\1.self_attn.z^layers\.(\d+)\.ffn\.zlayers.\1.mlp.z%^layers\.(\d+)\.self_attn\.attn_sink$zlayers.\1.self_attn.sinksz6^layers\.(\d+)\.self_attn\.indexer\.compressor\.norm\.z/layers.\1.self_attn.compressor.indexer.kv_norm.z4^layers\.(\d+)\.self_attn\.indexer\.compressor\.ape$z4layers.\1.self_attn.compressor.indexer.position_biasz0^layers\.(\d+)\.self_attn\.indexer\.compressor\.z'layers.\1.self_attn.compressor.indexer.z$^layers\.(\d+)\.self_attn\.indexer\.z-^layers\.(\d+)\.self_attn\.compressor\.norm\.z'layers.\1.self_attn.compressor.kv_norm.z+^layers\.(\d+)\.self_attn\.compressor\.ape$z,layers.\1.self_attn.compressor.position_biasz(^layers\.(\d+)\.self_attn\.(.*?)\.wq_a\.z layers.\1.self_attn.\2.q_a_proj.z(^layers\.(\d+)\.self_attn\.(.*?)\.wq_b\.z layers.\1.self_attn.\2.q_b_proj.z'^layers\.(\d+)\.self_attn\.(.*?)\.wkv\.zlayers.\1.self_attn.\2.kv_proj.z)^layers\.(\d+)\.self_attn\.(.*?)\.wgate\.z!layers.\1.self_attn.\2.gate_proj.z(^layers\.(\d+)\.self_attn\.(.*?)\.wo_a\.z layers.\1.self_attn.\2.o_a_proj.z(^layers\.(\d+)\.self_attn\.(.*?)\.wo_b\.z layers.\1.self_attn.\2.o_b_proj.z!^layers\.(\d+)\.self_attn\.wq_a\.zlayers.\1.self_attn.q_a_proj.z!^layers\.(\d+)\.self_attn\.wq_b\.zlayers.\1.self_attn.q_b_proj.z ^layers\.(\d+)\.self_attn\.wkv\.zlayers.\1.self_attn.kv_proj.z!^layers\.(\d+)\.self_attn\.wo_a\.zlayers.\1.self_attn.o_a_proj.z!^layers\.(\d+)\.self_attn\.wo_b\.zlayers.\1.self_attn.o_b_proj.z#^layers\.(\d+)\.self_attn\.q_norm\.zlayers.\1.self_attn.q_a_norm.z ^layers\.(\d+)\.mlp\.gate\.bias$z*layers.\1.mlp.gate.e_score_correction_biasz)^layers\.(\d+)\.mlp\.shared_experts\.w1\.z'layers.\1.mlp.shared_experts.gate_proj.z)^layers\.(\d+)\.mlp\.shared_experts\.w2\.z'layers.\1.mlp.shared_experts.down_proj.z)^layers\.(\d+)\.mlp\.shared_experts\.w3\.z%layers.\1.mlp.shared_experts.up_proj.zmlp.experts.*.w1.weightzmlp.experts.*.w3.weightzmlp.experts.gate_up_proj$r   zmlp.experts.*.w2.weightzmlp.experts.down_proj$rO   z^language_model.modellanguage_modelrC   z^language_model.lm_headlm_headz^language_modelzmodel.language_modelz^vision_towerzmodel.vision_towerz^multi_modal_projectorzmodel.multi_modal_projectorrL   z^image_newlinezmodel.image_newliner3   vision_model)prefix_to_remover=   
text_modelVideoLlavaModelvideo_llavaz^image_towerzmodel.image_towerz^video_towerzmodel.video_towerfuyuz^vision_embed_tokenszmodel.vision_embed_tokensmllamaz^vision_modelzmodel.vision_model	Emu3Modelz^text_model.modelemu3z^text_model.lm_headz^text_modelzmodel.text_modelz^vqmodelzmodel.vqmodelpaddleocr_vlz^mlp_ARzmodel.projectorz^visualzmodel.visualz1^model(?!(\.visual|\.projector|\.language_model))r_   z#^model(?!\.(language_model|visual))colqwen2modelvlm)r   model_prefixshieldgemma2)prefix_to_addr   timm_wrapper
timm_model)r   pi0
state_projzembed_action_time.state_projaction_in_projz embed_action_time.action_in_projaction_time_mlp_inz$embed_action_time.action_time_mlp_inaction_time_mlp_outz%embed_action_time.action_time_mlp_outz&^paligemma_with_expert.paligemma.modelz	model.vlmz)^paligemma_with_expert.gemma_expert.modelz	model.ditz+^paligemma_with_expert.gemma_expert.lm_headzmodel.dit.embed_tokensz(^paligemma_with_expert.paligemma.lm_headz%model.vlm.language_model.embed_tokensesmz9encoder.layer.*.attention.self.rotary_embeddings.inv_freqzrotary_embeddings.inv_freqdinov3_convnextz(?<!model\.)stageszmodel.stages
dinov3_vitz(?<!model\.)layer.zmodel.layer.
timesfm2_5ff0fc1ff1fc2olmo_hybridattention_layer_norminput_layernormfeedforward_layer_normpost_attention_layernormqwen3_5_textr1   z'detector_model.vision_encoder.backbone.zvision_encoder.backbone.ztracker_neck.zvision_encoder.neck.ztracker_model.(.+)z\1t5gemma2_encoderz-(?<!decoder\.)(?<!text_model\.)embed_tokens\.ztext_model.embed_tokens.z5(?<!decoder\.)(?<!text_model\.)(?<!layer)(?<!_)norm\.ztext_model.norm.zA(?<!vision_model.encoder\.)(?<!decoder\.)(?<!text_model\.)layers.ztext_model.layers.r   z.block_sparse_moe.z.mlp.z.experts.*.w1.weightz.experts.*.w3.weightz.experts.gate_up_projz.experts.*.w2.weightz.experts.down_projr   zmlp.experts.*.gate_proj.weightzmlp.experts.*.up_proj.weightzmlp.experts.gate_up_projzmlp.experts.*.down_proj.weightzmlp.experts.down_projqwen3_vl_moe   T)
check_dimsphimoez.gate.weightz.router.weightlfm2_moez feed_forward.experts.*.w1.weightz feed_forward.experts.*.w3.weightz!feed_forward.experts.gate_up_projz feed_forward.experts.*.w2.weightzfeed_forward.experts.down_projernie4_5_vl_moevision_towerzspatial_linear.0zspatial_linear.fc1zspatial_linear.2zspatial_linear.fc2zspatial_linear.3zspatial_linear.lnztemporal_linear.0ztemporal_linear.fc1ztemporal_linear.2ztemporal_linear.fc2ztemporal_linear.3ztemporal_linear.lnz!(?<!language_model\.)embed_tokenszlanguage_model.embed_tokensz(?<!language_model\.)layerszlanguage_model.layersz(?<!_)(?<!\w)norm\.zlanguage_model.norm.zmlp.gate.weight_1zmlp.vision_moe.gate.weight)dim0dim1zmlp.gate.weightzmlp.text_moe.gate.weightz'mlp.moe_statics.e_score_correction_biasz5mlp.text_moe.gate.moe_statics.e_score_correction_biasz7mlp.vision_moe.gate.moe_statics.e_score_correction_biaszexperts.*.down_proj.weightztext_moe.experts.down_projzvision_moe.experts.down_proj)	stack_dim
concat_dimzexperts.*.gate_proj.weightzexperts.*.up_proj.weightztext_moe.experts.gate_up_projzvision_moe.experts.gate_up_proj
maskformerzlayers.(\d+).fc1zlayers.\1.mlp.fc1zlayers.(\d+).fc2zlayers.\1.mlp.fc2r]   zbackbone.conv_encoderbackboneout_projo_projr-   zencoder.encoder.(\d+).layerszencoder.aifi.\1.layersRfDetrModelzbackbone.0.encoder.encoderzbackbone.backbonezbackbone.0.projectorzbackbone.projectorzprojector.stages.0.0.cv1.convz$projector.projector_layer.conv1.convzprojector.stages.0.0.cv1.bnz$projector.projector_layer.conv1.normzprojector.stages.0.0.cv2.convz$projector.projector_layer.conv2.convzprojector.stages.0.0.cv2.bnz$projector.projector_layer.conv2.normzprojector.stages.0.1zprojector.layer_normz%projector.stages.0.0.m.(\d+).cv1.convz3projector.projector_layer.bottlenecks.\1.conv1.convz#projector.stages.0.0.m.(\d+).cv1.bnz3projector.projector_layer.bottlenecks.\1.conv1.normz%projector.stages.0.0.m.(\d+).cv2.convz3projector.projector_layer.bottlenecks.\1.conv2.convz#projector.stages.0.0.m.(\d+).cv2.bnz3projector.projector_layer.bottlenecks.\1.conv2.normztransformer.decoderdecoderzdecoder.layers.(\d+).norm1z&decoder.layers.\1.self_attn_layer_normzdecoder.layers.(\d+).norm2z'decoder.layers.\1.cross_attn_layer_normzdecoder.layers.(\d+).linear1zdecoder.layers.\1.mlp.fc1zdecoder.layers.(\d+).linear2zdecoder.layers.\1.mlp.fc2zdecoder.layers.(\d+).norm3zdecoder.layers.\1.layer_normzdecoder.normzdecoder.layernormz^transformer\.enc_output_normenc_output_normz^transformer\.enc_output
enc_outputztransformer.enc_out_class_embedenc_out_class_embedztransformer.enc_out_bbox_embedenc_out_bbox_embedzrefpoint_embed\.weightzreference_point_embed.weightzself_attn.out_projzself_attn.o_projzself_attn.in_proj_bias)zself_attn.q_proj.biaszself_attn.k_proj.biaszself_attn.v_proj.biaszself_attn.in_proj_weight)rn   ro   rp   RfDetrForObjectDetection^zmodel.RfDetrForInstanceSegmentationz^(?!segmentation_head)z2segmentation_head\.query_features_block\.layers\.0zquery_features_block.mlp.fc1z2segmentation_head\.query_features_block\.layers\.2zquery_features_block.mlp.fc2z0segmentation_head\.query_features_block\.norm_inzquery_features_block.normz&segmentation_head\.blocks\.(\d+)\.normzblocks.\1.layernormz(segmentation_head\.blocks\.(\d+)\.dwconvzblocks.\1.depthwise_convz)segmentation_head\.blocks\.(\d+)\.pwconv1zblocks.\1.pointwise_convz(segmentation_head\.spatial_features_projspatial_features_projz&segmentation_head\.query_features_projquery_features_projzsegmentation_head\.biassegmentation_biasConditionalDetrModelzencoder_attn.out_projzencoder_attn.o_projz%decoder.layers.(\d+).sa_qcontent_projz*decoder.layers.\1.self_attn.q_content_projz!decoder.layers.(\d+).sa_qpos_projz&decoder.layers.\1.self_attn.q_pos_projz%decoder.layers.(\d+).sa_kcontent_projz*decoder.layers.\1.self_attn.k_content_projz!decoder.layers.(\d+).sa_kpos_projz&decoder.layers.\1.self_attn.k_pos_projzdecoder.layers.(\d+).sa_v_projz"decoder.layers.\1.self_attn.v_projz%decoder.layers.(\d+).ca_qcontent_projz-decoder.layers.\1.encoder_attn.q_content_projz!decoder.layers.(\d+).ca_qpos_projz)decoder.layers.\1.encoder_attn.q_pos_projz%decoder.layers.(\d+).ca_kcontent_projz-decoder.layers.\1.encoder_attn.k_content_projz!decoder.layers.(\d+).ca_kpos_projz)decoder.layers.\1.encoder_attn.k_pos_projzdecoder.layers.(\d+).ca_v_projz%decoder.layers.\1.encoder_attn.v_projz&decoder.layers.(\d+).ca_qpos_sine_projz.decoder.layers.\1.encoder_attn.q_pos_sine_projzlayers.\1.mlp.layers.0zlayers.\1.mlp.layers.1z	backbone.zembedding.weightzembeddings.weightzmixer.experts.*.up_proj.weightzmixer.experts.up_projz mixer.experts.*.down_proj.weightzmixer.experts.down_projz'feed_forward.experts.*.gate_proj.weightz%feed_forward.experts.*.up_proj.weightz'feed_forward.experts.*.down_proj.weightzencoder.layerslayersemb_lnzembeddings.LayerNormzattn.out_projfc11up_projfc12	gate_proj	down_projnorm1norm2post_mlp_layernormz	attn.Wqkv)zself_attn.q_projzself_attn.k_projzself_attn.v_projzmixer.out_projz
mixer.Wqkvzencoder\.pre_encode\.conv\.zencoder.subsampling.layers.zencoder\.pre_encode\.out\.zencoder.subsampling.linear.z7transf_decoder\._embedding\.position_embedding\.pos_enczdecoder.pos_emb.weightz+transf_decoder\._embedding\.token_embeddingzdecoder.embed_tokensz&transf_decoder\._embedding\.layer_normzdecoder.embedding_layernormz*transf_decoder\._decoder\.final_layer_normz transf_decoder\._decoder\.layerszdecoder.layerszencoder_decoder_proj\.zdecoder.proj.z"encoder\.(.+)\.self_attn\.linear_qzencoder.\1.self_attn.q_projz"encoder\.(.+)\.self_attn\.linear_kzencoder.\1.self_attn.k_projz"encoder\.(.+)\.self_attn\.linear_vzencoder.\1.self_attn.v_projz$encoder\.(.+)\.self_attn\.linear_outzencoder.\1.self_attn.o_projz$encoder\.(.+)\.self_attn\.linear_posz$encoder.\1.self_attn.relative_k_projz$encoder\.(.+)\.self_attn\.pos_bias_uzencoder.\1.self_attn.bias_uz$encoder\.(.+)\.self_attn\.pos_bias_vzencoder.\1.self_attn.bias_vz)decoder\.(.+)\.first_sub_layer\.query_netzdecoder.\1.self_attn.q_projz'decoder\.(.+)\.first_sub_layer\.key_netzdecoder.\1.self_attn.k_projz)decoder\.(.+)\.first_sub_layer\.value_netzdecoder.\1.self_attn.v_projz.decoder\.(.+)\.first_sub_layer\.out_projectionzdecoder.\1.self_attn.o_projz\.second_sub_layer\.query_netz.encoder_attn.q_projz\.second_sub_layer\.key_netz.encoder_attn.k_projz\.second_sub_layer\.value_netz.encoder_attn.v_projz"\.second_sub_layer\.out_projectionz.encoder_attn.o_projz\.third_sub_layer\.dense_inz.mlp.fc1z\.third_sub_layer\.dense_outz.mlp.fc2z\.layer_norm_1\.z.input_layernorm.z\.layer_norm_2\.z.post_attention_layernorm.z\.layer_norm_3\.z.final_layernorm.z\.conv\.batch_normz
.conv.normzlog_softmax\.mlp\.layer0proj_outz^vision_model\.zmodel\.vision_tower\.zencoder\.layers\.zlayers\.z\.ls1z
\.lambda_1z\.ls2z
\.lambda_2z(layers\.\d+)\.attn\.proj\.z\1.attention.projection_layer.z	\.norm1\.z\.layernorm_before\.z	\.norm2\.z\.layernorm_after\.z\.embeddings\.class_embeddingz\.embeddings\.cls_tokenz \.embeddings\.position_embeddingz!\.embeddings\.position_embeddingsz\.embeddings\.patch_embedding\.z,\.embeddings\.patch_embeddings\.projection\.z^language_model\.model\.zmodel\.language_model\.z^language_model\.lm_head\.z	lm_head\.z
^mlp1\.0\.z*model\.multi_modal_projector\.layer_norm\.z
^mlp1\.1\.z(model\.multi_modal_projector\.linear_1\.z
^mlp1\.3\.z(model\.multi_modal_projector\.linear_2\.zattn.qkv.weight)zattention.q_proj.weightzattention.k_proj.weightzattention.v_proj.weightzattn.qkv.bias)zattention.q_proj.biaszattention.k_proj.biaszattention.v_proj.biaszLayerNorm.gammazLayerNorm.weightzLayerNorm.betazLayerNorm.bias)	deformable_detrd_fine
nemotron_hjamba
nomic_bertjina_embeddings_v3
cohere_asrqianfan_ocrlegacyr   z
.weight_g$z".parametrizations.weight.original0z
.weight_v$z".parametrizations.weight.original1zbbox_attention.q_linearzbbox_attention.q_projzbbox_attention.k_linearzbbox_attention.k_projzmask_head.lay1zmask_head.conv1.convzmask_head.gn1zmask_head.conv1.normzmask_head.lay2zmask_head.conv2.convzmask_head.gn2zmask_head.conv2.normzmask_head.adapter1z"mask_head.fpn_stages.0.fpn_adapterzmask_head.lay3z"mask_head.fpn_stages.0.refine.convzmask_head.gn3z"mask_head.fpn_stages.0.refine.normzmask_head.adapter2z"mask_head.fpn_stages.1.fpn_adapterzmask_head.lay4z"mask_head.fpn_stages.1.refine.convzmask_head.gn4z"mask_head.fpn_stages.1.refine.normzmask_head.adapter3z"mask_head.fpn_stages.2.fpn_adapterzmask_head.lay5z"mask_head.fpn_stages.2.refine.convzmask_head.gn5z"mask_head.fpn_stages.2.refine.normzmask_head.out_layzmask_head.output_convDetrForSegmentationConditionalDetrForSegmentationr   z,mlp.gate.moe_statics.e_score_correction_biaszencoder.relative_position_biasshared_position_biaszGattention.attention.relative_position_bias.relative_position_bias_tablez3relative_position_bias.relative_position_bias_tablerb   pixiozpixio.r   z).block_sparse_moe.e_score_correction_biasz.mlp.e_score_correction_biasr*   zmlp.e_score_correction_biasz mlp.gate.e_score_correction_biashy_v3zmlp\.router\.gate\.weightzmlp\.expert_biaszmlp\.shared_mlp\.zmlp.shared_experts.qwen3_5_moe_textlagunaz#mlp.experts.e_score_correction_biaszmlp.shared_expert.)r   r   r   r   r	   r   r   r
   copy_MODEL_TO_CONVERSION_PATTERNitems)mapping
model_typebase_patterns      _/var/www/html/banglarbhumi/venv/lib/python3.11/site-packages/transformers/conversion_mapping.py$_build_checkpoint_conversion_mappingr   w   s#   f 9!79M N!aLLL>  
  7! ! ! "aLLL>	 	 	 +>Pdeee!
f& 	.	::,h77?H55,h7735GHH/;;>955
'f8 	'=.99)
9f> 	,h77?H55,h7735GHH/;;>955!
?fN 	=*EFF=*CDD=*EFF=+66/1FGG+-?@@,h77?H55,h7735GHH/;;>955
Ofj 	&>@PQQFH\]]=*EFF=*CDD=*EFF=+6646STTCL  ;H (
kfL 	68JKK46HII68JKK-/ABB	
MfX 	=?]^^24HII68OPP13EFF/1CDD13EFF.0abb68abb35GHH<33<33>+=>>>+<==
Yfv 	+^<RTt-u-u,vwfx 	13EFF/1CDD13EFF=O  35GHH/;;>955
yfP 	=/::+-?@@
QfX 	:yQQQ
Yf^ 	 v
6 +=Odeee7v
8 +<N^___9v
: +<m\\\;v
< ?O\\\=v
> +<N_```?v
@ +=OabbbAv
B  > =  Cv
J  = F  Kv
R +I[rsssSv
T  @Rk  Uv
Z  ASm  [v
` +HZpqqqav
b +J\tuuucv
d  @Rk  ev
j  9 7  kv
r  8 1  sv
D  H <  Ev
L  Y R  Mv
T  W W  Uv
\  S J  ]v
d  G J  ev
l  P J  mv
t  N O  uv
F  K C  Gv
N  K C  Ov
V  J B  Wv
^  L D  _v
f  K C  gv
n  K C  ov
v  D @  wv
~  D @  v
F  C ?  Gv
N  D @  Ov
V  D @  Wv
f  F @  gv
r  C M  sv
z  L J  {v
B  L J  Cv
J  L H  Kv
R --! !<+222KA4F4F4FG  Sv
b  9 8+2223  cv
_fL	 	+CUefff
M	fR	 	+EW`aaa+=Oefff+;Mabbb+DVsttt	
S	f^	 	+EW`aaa+=Oefff+;Mabbb+DVsttt+<Ncddd
_	fl	 	lNKKKLm	f fn	 	L,GGGHo	fp	 	+CUefff
q	fv	 	+EW`aaa+=Oefff?L_```?L_```+DVsttt
w	fD
 	+EW`aaa+=Oefff+BToppp
E
fN
 	+EW`aaa+=Oefff+;Mabbb+DVsttt	
O
fZ
 	+?Q]^^^
[
f`
 	+AS\]]]>K]^^^;XXX
a
fj
 	:GXYYY:~VVV T 6  
k
fz
 	*:~VVV FXn  ,
{
fF 	\7OOOPGfH 	G'RRRSIfJ 	LAAABKfL 	=Jhiii+<Npqqq 5Gm    6Ho   +Tfqrrr+Wituuu N 8    K G  !
Mfv 	K, 
wfB 	N+@/RRSCfD 	~&;_MMNEfF 	5%((5%((
Gf f fN 	13DEE35OPP
OfV 	7GV]^^^_WfX 	 J\v   +;Mcddd+@RWXXX
Yfh 	KMghhSUghh_auvv
ifr 	/99**! !8#    A&&&	   *! !5#  
 
 

sfh 	42! !;+222KA4F4F4FG    @ 7+2223  
ifF 	 : :%at<<<=  
  7 7%at<<<=  
Gf^ 	/99>+;<<**! !8+222KA4F4F4FG    6 4+2223  
_f@ 	66! !D+222KA4F4F4FG    B @+2223  
Af^ 	>>::-/CDD-/CDD-/BCC.0EFF.0EFF.0DEE?A^__9;RSS13IJJ 3 <%115556  
  1 :%115556  
 !J KKM! "aLLL>   != >02! ?WXYYYZ   0.!
 45! ?WXYYYZ
 
 
Q3
_fT 	.0DEE.0DEE
Uf\ 	2J??:x00.0DEE.0DEE	
]fh 	:x00.0DEE.0DEE:<UVV	
ift 	8:NOO24IJJ;=dee9;bcc;=dee9;bcc24KLL8:p  68n  8:p  68n  1:>>8:cdd8:dee:<XYY:<XYY8:YZZ?,@AA;=OPP6FF=?UVV<>STT46UVV02EFF 9 n n n!aLLL>  
  ; t t t!aLLL>  S.
ufR 	#3III%
SfX 	(+CU]^^^PRqrrPRqrrNPlmmDF\]]FHcddGIdeeFH`aaDF\]]57KLL*
Yfp 	2J??/1CDD24IJJ.0DEE.0DEECErss?AjkkCErss?Ajkk<>cddCEuvv?AmnnCEuvv?Amnn<>fgg9;l %!
qf f^ 2J??/1CDD.0DEE.0DEE	
 :x00.0IJJ.0IJJ:<UVV	
 ;11-/BCC4! !8+2223   6! !:+2223  
& =;! !D+222KA4F4F4FG    I @+2223  
  ,i889&=>>+-@AA7J//7L116<008%@AA%  !,! ! !
 "aLLL>  
. 8E[\\\+;XVVV+;M_```7D^___7DXYYY ,! ! !
 "aLLL>  
" 9;YZZ8:XYYUWpqqIKbccDFdeeH/ZZ>@QRR46FGG@B`aa@B`aa@B`aaBDbccBDkllBDbccBDbccGIghhEGeffGIghhLNlmm;=TUU9;RSS;=TUU@BYZZ9;GG:KHH.0DEE.0MNN.0DEE0-@@6DD=
B -/GHH/==8]338]339;\]]<)@AA<)?@@;=WXX>@dee=?noo68RSS8,GG=*WXX=*UVV=*UVV!2 3 q q q!aLLL>  
 !0 1 k k k!aLLL>  +
8  1 2    0 0  	
yf f fGP ENNNgh',,...NNNH(@	
 	
 	
 	(@	
 	
 	
	 	 	02IJJ02IJJ')?@@(>??')?@@(>??+-QRR')MNN(LMM+-QRR')MNN(LMM+-QRR')MNN(LMM*,CDD!&G!"$ 188M0N0S0S0U0UG,-%k27799GNN@Bpqq   #:.335579OPPUA	
 	
9 GK z*//11GGGw 233w 122}h//  $I.3355GLLBDbcc  $K05577GLLn-JLnoopp {+0022GGG'CUfggg':Lijjj';Mbccc 
 #*."9">">"@"@G7;#7#<#<#>#>>,1133GHH<>`aa+-BCC 
 %A$F$F$H$H ; ; 
L  %l388::
N    Nc                z    t           t                      a t          t                               |                     S )N)$_checkpoint_conversion_mapping_cacher   r   get)r   s    r   !get_checkpoint_conversion_mappingr     s1    +3/S/U/U,8<<ZHHIIIr   Fmodel_type_or_class_namestrr   &list[WeightConverter | WeightRenaming]	overwriteboolreturnNonec                    t           t                      a | t           v r|st          d|  d          |t           | <   dS )aR  
    Register a conversion mapping for a model type string or a class name.

    Class names take priority over `model_type` strings during lookup (see
    `extract_weight_conversions_for_model`), making it possible to define
    task-head-specific or class-specific conversions that differ from the shared
    `model_type` baseline.
    NzConversion mapping for 'z4' already exists. Pass overwrite=True to replace it.)r   r   
ValueError)r   r   r   s      r   &register_checkpoint_conversion_mappingr     s_     ,3/S/U/U,#GGGPYGu'?uuu
 
 	
 FM()ABBBr   r   r   list[WeightTransform] | Nonec                    t          |           j        }| j        j        }t	          |          }||rt	          |          }|S )aF  
    Return the registered conversion list for `model`, or `None` if none exists.

    Looks up by class name first (enables task-head-specific overrides), then
    falls back to `model.config.model_type`.  Transforms are returned
    unmodified; the caller sets `scope_prefix` on each transform for sub-module isolation.
    )type__name__configr   r   )r   
class_namer   conversionss       r   $extract_weight_conversions_for_modelr    sI     e%J(J 4J??Kz7
CCr   Tkey_mappingdict[str, str] | Nonehf_quantizerHfQuantizer | None
add_legacylist[WeightTransform]c                4   ddl m} g }|d |                                D             }t          t                    }|                                 D ]	\  }t          ||          st          |          j        }|j	        j
        }	t          fd||         D                       rYt          |          }
|
du}|s5|	r"t          fd||	         D                       r|	t          |	          }
|
dk    }|s|
D ]	}|_        
|                    |
           ||                                        |s|	r||	                                        |r"|                    t          d                     ||                    |          }|S )	a-  
    Collect the ordered list of weight transforms for `model` (used during
    loading and, when reversed, during saving).

    Each `PreTrainedModel` sub-module is looked up by class name then
    `model_type`.  Root transforms are applied globally; sub-module transforms
    have their `scope_prefix` set so they only match keys under that prefix.  After any
    sub-module is processed, both its class name and `model_type` are marked
    seen to prevent `XForY` / `XModel` pairs from applying the same mapping
    twice via different lookup paths.
    r   r   Nc                6    g | ]\  }}t          ||           S )rq   )r   ).0kvs      r   
<listcomp>z0get_model_conversion_mapping.<locals>.<listcomp>A  s-    sssW[WXZ[nQPQRRRsssr   c              3  T   K   | ]"}|d k    p                     |dz             V  #dS  .N
startswithr
  seenmodule_names     r   	<genexpr>z/get_model_conversion_mapping.<locals>.<genexpr>T  s>      iiDtrz?[33D3J??iiiiiir   c              3  T   K   | ]"}|d k    p                     |dz             V  #dS r  r  r  s     r   r  z/get_model_conversion_mapping.<locals>.<genexpr>_  sO       " "EI
@k44TCZ@@" " " " " "r   r  r   )modeling_utilsr   r   r   listnamed_modules
isinstancer   r   r   r   anyr   scope_prefixextendappendupdate_weight_conversions)r   r  r  r  r   weight_conversionsseen_identifiers	submoduler   r   r   found_via_classis_root_model	transformr  s                 @r   get_model_conversion_mappingr(  )  s-   " 0/////  ss_j_p_p_r_rsss 5@4E4E"'"5"5"7"7 += +=Y)_55 	)__-
%0
 iiiiL\]gLhiiiii 	
 8
CC%T1 	L c " " " "M]^hMi" " "    %?
KK#r) 	5( 5 5	)4	&&!!+...$++K888
  	=: 	=Z(//<<< O!!"CH"M"MNNN )CCDVWWr   )F)r   r   r   r   r   r   r   r   )r   r   r   r   )NNT)
r   r   r  r  r  r  r  r   r   r  )
__future__r   collectionsr   r   r   typingr   core_model_loadingr   r	   r
   r   r   r   r   r   r   r  r   
quantizersr   r   r   r   r   r   r  r(   r   r   <module>r/     s   # " " " " " # # # # # #                  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  (//////''''''M yM  )M 
 [M  ;M  ;M  [M  KM  M  [M  ;M  M  [M  +M   !M " k#M $ k%M & +'M  M ( k)M * +M , [-M . +/M 0 ;1M 2 )3M 4 y5M 6 y7M 8 .9M : /;M <  !4=M > /?M @ 2AM B &CM D EM F .GM H /IM  M  M J -KM L +MM N .OM P *QM R +SM T )UM V WM X 'YM Z [M \ g]M ^ _M ` aM b cM d geM f gM h |iM l lmM  M  M n loM p LqM r <sM t \uM v \wM x \yM z ,{M | l}M ~ <M @ <AM B CM D <EM F [GM H )*KIM L 
MM N OM P QM  M R YM  M  M  `z z zz (, $J J J M M M M M.   , *.'+	Y Y Y Y Y Y Yr   