modelscope/DiffSynth-Painter

1import torch2from transformers import T5EncoderModel, T5Config3from .sd_text_encoder import SDTextEncoder4from .sdxl_text_encoder import SDXLTextEncoder2, SDXLTextEncoder2StateDictConverter5    6 7class SD3TextEncoder1(SDTextEncoder):8    def __init__(self, vocab_size=49408):9        super().__init__(vocab_size=vocab_size)10 11    def forward(self, input_ids, clip_skip=2):12        embeds = self.token_embedding(input_ids) + self.position_embeds13        attn_mask = self.attn_mask.to(device=embeds.device, dtype=embeds.dtype)14        for encoder_id, encoder in enumerate(self.encoders):15            embeds = encoder(embeds, attn_mask=attn_mask)16            if encoder_id + clip_skip == len(self.encoders):17                hidden_states = embeds18        embeds = self.final_layer_norm(embeds)19        pooled_embeds = embeds[torch.arange(embeds.shape[0]), input_ids.to(dtype=torch.int).argmax(dim=-1)]20        return pooled_embeds, hidden_states21    22    @staticmethod23    def state_dict_converter():24        return SD3TextEncoder1StateDictConverter()25    26 27 28class SD3TextEncoder2(SDXLTextEncoder2):29    def __init__(self):30        super().__init__()31 32    @staticmethod33    def state_dict_converter():34        return SD3TextEncoder2StateDictConverter()35 36 37class SD3TextEncoder3(T5EncoderModel):38    def __init__(self):39        config = T5Config(40            _name_or_path = ".",41            architectures = ["T5EncoderModel"],42            classifier_dropout = 0.0,43            d_ff = 10240,44            d_kv = 64,45            d_model = 4096,46            decoder_start_token_id = 0,47            dense_act_fn = "gelu_new",48            dropout_rate = 0.1,49            eos_token_id = 1,50            feed_forward_proj = "gated-gelu",51            initializer_factor = 1.0,52            is_encoder_decoder = True,53            is_gated_act = True,54            layer_norm_epsilon = 1e-06,55            model_type = "t5",56            num_decoder_layers = 24,57            num_heads = 64,58            num_layers = 24,59            output_past = True,60            pad_token_id = 0,61            relative_attention_max_distance = 128,62            relative_attention_num_buckets = 32,63            tie_word_embeddings = False,64            torch_dtype = torch.float16,65            transformers_version = "4.41.2",66            use_cache = True,67            vocab_size = 3212868        )69        super().__init__(config)70        self.eval()71 72    def forward(self, input_ids):73        outputs = super().forward(input_ids=input_ids)74        prompt_emb = outputs.last_hidden_state75        return prompt_emb76    77    @staticmethod78    def state_dict_converter():79        return SD3TextEncoder3StateDictConverter()80 81 82 83class SD3TextEncoder1StateDictConverter:84    def __init__(self):85        pass86 87    def from_diffusers(self, state_dict):88        rename_dict = {89            "text_model.embeddings.token_embedding.weight": "token_embedding.weight",90            "text_model.embeddings.position_embedding.weight": "position_embeds",91            "text_model.final_layer_norm.weight": "final_layer_norm.weight",92            "text_model.final_layer_norm.bias": "final_layer_norm.bias",93        }94        attn_rename_dict = {95            "self_attn.q_proj": "attn.to_q",96            "self_attn.k_proj": "attn.to_k",97            "self_attn.v_proj": "attn.to_v",98            "self_attn.out_proj": "attn.to_out",99            "layer_norm1": "layer_norm1",100            "layer_norm2": "layer_norm2",101            "mlp.fc1": "fc1",102            "mlp.fc2": "fc2",103        }104        state_dict_ = {}105        for name in state_dict:106            if name in rename_dict:107                param = state_dict[name]108                if name == "text_model.embeddings.position_embedding.weight":109                    param = param.reshape((1, param.shape[0], param.shape[1]))110                state_dict_[rename_dict[name]] = param111            elif name.startswith("text_model.encoder.layers."):112                param = state_dict[name]113                names = name.split(".")114                layer_id, layer_type, tail = names[3], ".".join(names[4:-1]), names[-1]115                name_ = ".".join(["encoders", layer_id, attn_rename_dict[layer_type], tail])116                state_dict_[name_] = param117        return state_dict_118    119    def from_civitai(self, state_dict):120        rename_dict = {121            "text_encoders.clip_l.transformer.text_model.embeddings.position_embedding.weight": "position_embeds",122            "text_encoders.clip_l.transformer.text_model.embeddings.token_embedding.weight": "token_embedding.weight",123            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.layer_norm1.bias": "encoders.0.layer_norm1.bias",124            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.layer_norm1.weight": "encoders.0.layer_norm1.weight",125            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.layer_norm2.bias": "encoders.0.layer_norm2.bias",126            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.layer_norm2.weight": "encoders.0.layer_norm2.weight",127            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.mlp.fc1.bias": "encoders.0.fc1.bias",128            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.mlp.fc1.weight": "encoders.0.fc1.weight",129            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.mlp.fc2.bias": "encoders.0.fc2.bias",130            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.mlp.fc2.weight": "encoders.0.fc2.weight",131            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.k_proj.bias": "encoders.0.attn.to_k.bias",132            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.k_proj.weight": "encoders.0.attn.to_k.weight",133            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.out_proj.bias": "encoders.0.attn.to_out.bias",134            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.out_proj.weight": "encoders.0.attn.to_out.weight",135            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.q_proj.bias": "encoders.0.attn.to_q.bias",136            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.q_proj.weight": "encoders.0.attn.to_q.weight",137            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.v_proj.bias": "encoders.0.attn.to_v.bias",138            "text_encoders.clip_l.transformer.text_model.encoder.layers.0.self_attn.v_proj.weight": "encoders.0.attn.to_v.weight",139            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.layer_norm1.bias": "encoders.1.layer_norm1.bias",140            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.layer_norm1.weight": "encoders.1.layer_norm1.weight",141            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.layer_norm2.bias": "encoders.1.layer_norm2.bias",142            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.layer_norm2.weight": "encoders.1.layer_norm2.weight",143            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.mlp.fc1.bias": "encoders.1.fc1.bias",144            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.mlp.fc1.weight": "encoders.1.fc1.weight",145            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.mlp.fc2.bias": "encoders.1.fc2.bias",146            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.mlp.fc2.weight": "encoders.1.fc2.weight",147            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.k_proj.bias": "encoders.1.attn.to_k.bias",148            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.k_proj.weight": "encoders.1.attn.to_k.weight",149            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.out_proj.bias": "encoders.1.attn.to_out.bias",150            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.out_proj.weight": "encoders.1.attn.to_out.weight",151            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.q_proj.bias": "encoders.1.attn.to_q.bias",152            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.q_proj.weight": "encoders.1.attn.to_q.weight",153            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.v_proj.bias": "encoders.1.attn.to_v.bias",154            "text_encoders.clip_l.transformer.text_model.encoder.layers.1.self_attn.v_proj.weight": "encoders.1.attn.to_v.weight",155            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.layer_norm1.bias": "encoders.10.layer_norm1.bias",156            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.layer_norm1.weight": "encoders.10.layer_norm1.weight",157            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.layer_norm2.bias": "encoders.10.layer_norm2.bias",158            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.layer_norm2.weight": "encoders.10.layer_norm2.weight",159            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.mlp.fc1.bias": "encoders.10.fc1.bias",160            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.mlp.fc1.weight": "encoders.10.fc1.weight",161            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.mlp.fc2.bias": "encoders.10.fc2.bias",162            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.mlp.fc2.weight": "encoders.10.fc2.weight",163            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.k_proj.bias": "encoders.10.attn.to_k.bias",164            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.k_proj.weight": "encoders.10.attn.to_k.weight",165            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.out_proj.bias": "encoders.10.attn.to_out.bias",166            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.out_proj.weight": "encoders.10.attn.to_out.weight",167            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.q_proj.bias": "encoders.10.attn.to_q.bias",168            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.q_proj.weight": "encoders.10.attn.to_q.weight",169            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.v_proj.bias": "encoders.10.attn.to_v.bias",170            "text_encoders.clip_l.transformer.text_model.encoder.layers.10.self_attn.v_proj.weight": "encoders.10.attn.to_v.weight",171            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.layer_norm1.bias": "encoders.11.layer_norm1.bias",172            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.layer_norm1.weight": "encoders.11.layer_norm1.weight",173            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.layer_norm2.bias": "encoders.11.layer_norm2.bias",174            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.layer_norm2.weight": "encoders.11.layer_norm2.weight",175            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.mlp.fc1.bias": "encoders.11.fc1.bias",176            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.mlp.fc1.weight": "encoders.11.fc1.weight",177            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.mlp.fc2.bias": "encoders.11.fc2.bias",178            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.mlp.fc2.weight": "encoders.11.fc2.weight",179            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.k_proj.bias": "encoders.11.attn.to_k.bias",180            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.k_proj.weight": "encoders.11.attn.to_k.weight",181            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.out_proj.bias": "encoders.11.attn.to_out.bias",182            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.out_proj.weight": "encoders.11.attn.to_out.weight",183            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.q_proj.bias": "encoders.11.attn.to_q.bias",184            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.q_proj.weight": "encoders.11.attn.to_q.weight",185            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.v_proj.bias": "encoders.11.attn.to_v.bias",186            "text_encoders.clip_l.transformer.text_model.encoder.layers.11.self_attn.v_proj.weight": "encoders.11.attn.to_v.weight",187            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.layer_norm1.bias": "encoders.2.layer_norm1.bias",188            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.layer_norm1.weight": "encoders.2.layer_norm1.weight",189            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.layer_norm2.bias": "encoders.2.layer_norm2.bias",190            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.layer_norm2.weight": "encoders.2.layer_norm2.weight",191            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.mlp.fc1.bias": "encoders.2.fc1.bias",192            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.mlp.fc1.weight": "encoders.2.fc1.weight",193            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.mlp.fc2.bias": "encoders.2.fc2.bias",194            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.mlp.fc2.weight": "encoders.2.fc2.weight",195            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.k_proj.bias": "encoders.2.attn.to_k.bias",196            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.k_proj.weight": "encoders.2.attn.to_k.weight",197            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.out_proj.bias": "encoders.2.attn.to_out.bias",198            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.out_proj.weight": "encoders.2.attn.to_out.weight",199            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.q_proj.bias": "encoders.2.attn.to_q.bias",200            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.q_proj.weight": "encoders.2.attn.to_q.weight",201            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.v_proj.bias": "encoders.2.attn.to_v.bias",202            "text_encoders.clip_l.transformer.text_model.encoder.layers.2.self_attn.v_proj.weight": "encoders.2.attn.to_v.weight",203            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.layer_norm1.bias": "encoders.3.layer_norm1.bias",204            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.layer_norm1.weight": "encoders.3.layer_norm1.weight",205            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.layer_norm2.bias": "encoders.3.layer_norm2.bias",206            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.layer_norm2.weight": "encoders.3.layer_norm2.weight",207            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.mlp.fc1.bias": "encoders.3.fc1.bias",208            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.mlp.fc1.weight": "encoders.3.fc1.weight",209            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.mlp.fc2.bias": "encoders.3.fc2.bias",210            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.mlp.fc2.weight": "encoders.3.fc2.weight",211            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.k_proj.bias": "encoders.3.attn.to_k.bias",212            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.k_proj.weight": "encoders.3.attn.to_k.weight",213            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.out_proj.bias": "encoders.3.attn.to_out.bias",214            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.out_proj.weight": "encoders.3.attn.to_out.weight",215            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.q_proj.bias": "encoders.3.attn.to_q.bias",216            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.q_proj.weight": "encoders.3.attn.to_q.weight",217            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.v_proj.bias": "encoders.3.attn.to_v.bias",218            "text_encoders.clip_l.transformer.text_model.encoder.layers.3.self_attn.v_proj.weight": "encoders.3.attn.to_v.weight",219            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.layer_norm1.bias": "encoders.4.layer_norm1.bias",220            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.layer_norm1.weight": "encoders.4.layer_norm1.weight",221            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.layer_norm2.bias": "encoders.4.layer_norm2.bias",222            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.layer_norm2.weight": "encoders.4.layer_norm2.weight",223            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.mlp.fc1.bias": "encoders.4.fc1.bias",224            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.mlp.fc1.weight": "encoders.4.fc1.weight",225            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.mlp.fc2.bias": "encoders.4.fc2.bias",226            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.mlp.fc2.weight": "encoders.4.fc2.weight",227            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.k_proj.bias": "encoders.4.attn.to_k.bias",228            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.k_proj.weight": "encoders.4.attn.to_k.weight",229            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.out_proj.bias": "encoders.4.attn.to_out.bias",230            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.out_proj.weight": "encoders.4.attn.to_out.weight",231            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.q_proj.bias": "encoders.4.attn.to_q.bias",232            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.q_proj.weight": "encoders.4.attn.to_q.weight",233            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.v_proj.bias": "encoders.4.attn.to_v.bias",234            "text_encoders.clip_l.transformer.text_model.encoder.layers.4.self_attn.v_proj.weight": "encoders.4.attn.to_v.weight",235            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.layer_norm1.bias": "encoders.5.layer_norm1.bias",236            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.layer_norm1.weight": "encoders.5.layer_norm1.weight",237            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.layer_norm2.bias": "encoders.5.layer_norm2.bias",238            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.layer_norm2.weight": "encoders.5.layer_norm2.weight",239            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.mlp.fc1.bias": "encoders.5.fc1.bias",240            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.mlp.fc1.weight": "encoders.5.fc1.weight",241            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.mlp.fc2.bias": "encoders.5.fc2.bias",242            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.mlp.fc2.weight": "encoders.5.fc2.weight",243            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.k_proj.bias": "encoders.5.attn.to_k.bias",244            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.k_proj.weight": "encoders.5.attn.to_k.weight",245            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.out_proj.bias": "encoders.5.attn.to_out.bias",246            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.out_proj.weight": "encoders.5.attn.to_out.weight",247            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.q_proj.bias": "encoders.5.attn.to_q.bias",248            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.q_proj.weight": "encoders.5.attn.to_q.weight",249            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.v_proj.bias": "encoders.5.attn.to_v.bias",250            "text_encoders.clip_l.transformer.text_model.encoder.layers.5.self_attn.v_proj.weight": "encoders.5.attn.to_v.weight",251            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.layer_norm1.bias": "encoders.6.layer_norm1.bias",252            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.layer_norm1.weight": "encoders.6.layer_norm1.weight",253            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.layer_norm2.bias": "encoders.6.layer_norm2.bias",254            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.layer_norm2.weight": "encoders.6.layer_norm2.weight",255            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.mlp.fc1.bias": "encoders.6.fc1.bias",256            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.mlp.fc1.weight": "encoders.6.fc1.weight",257            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.mlp.fc2.bias": "encoders.6.fc2.bias",258            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.mlp.fc2.weight": "encoders.6.fc2.weight",259            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.k_proj.bias": "encoders.6.attn.to_k.bias",260            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.k_proj.weight": "encoders.6.attn.to_k.weight",261            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.out_proj.bias": "encoders.6.attn.to_out.bias",262            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.out_proj.weight": "encoders.6.attn.to_out.weight",263            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.q_proj.bias": "encoders.6.attn.to_q.bias",264            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.q_proj.weight": "encoders.6.attn.to_q.weight",265            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.v_proj.bias": "encoders.6.attn.to_v.bias",266            "text_encoders.clip_l.transformer.text_model.encoder.layers.6.self_attn.v_proj.weight": "encoders.6.attn.to_v.weight",267            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.layer_norm1.bias": "encoders.7.layer_norm1.bias",268            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.layer_norm1.weight": "encoders.7.layer_norm1.weight",269            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.layer_norm2.bias": "encoders.7.layer_norm2.bias",270            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.layer_norm2.weight": "encoders.7.layer_norm2.weight",271            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.mlp.fc1.bias": "encoders.7.fc1.bias",272            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.mlp.fc1.weight": "encoders.7.fc1.weight",273            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.mlp.fc2.bias": "encoders.7.fc2.bias",274            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.mlp.fc2.weight": "encoders.7.fc2.weight",275            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.k_proj.bias": "encoders.7.attn.to_k.bias",276            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.k_proj.weight": "encoders.7.attn.to_k.weight",277            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.out_proj.bias": "encoders.7.attn.to_out.bias",278            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.out_proj.weight": "encoders.7.attn.to_out.weight",279            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.q_proj.bias": "encoders.7.attn.to_q.bias",280            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.q_proj.weight": "encoders.7.attn.to_q.weight",281            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.v_proj.bias": "encoders.7.attn.to_v.bias",282            "text_encoders.clip_l.transformer.text_model.encoder.layers.7.self_attn.v_proj.weight": "encoders.7.attn.to_v.weight",283            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.layer_norm1.bias": "encoders.8.layer_norm1.bias",284            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.layer_norm1.weight": "encoders.8.layer_norm1.weight",285            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.layer_norm2.bias": "encoders.8.layer_norm2.bias",286            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.layer_norm2.weight": "encoders.8.layer_norm2.weight",287            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.mlp.fc1.bias": "encoders.8.fc1.bias",288            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.mlp.fc1.weight": "encoders.8.fc1.weight",289            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.mlp.fc2.bias": "encoders.8.fc2.bias",290            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.mlp.fc2.weight": "encoders.8.fc2.weight",291            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.k_proj.bias": "encoders.8.attn.to_k.bias",292            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.k_proj.weight": "encoders.8.attn.to_k.weight",293            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.out_proj.bias": "encoders.8.attn.to_out.bias",294            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.out_proj.weight": "encoders.8.attn.to_out.weight",295            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.q_proj.bias": "encoders.8.attn.to_q.bias",296            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.q_proj.weight": "encoders.8.attn.to_q.weight",297            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.v_proj.bias": "encoders.8.attn.to_v.bias",298            "text_encoders.clip_l.transformer.text_model.encoder.layers.8.self_attn.v_proj.weight": "encoders.8.attn.to_v.weight",299            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.layer_norm1.bias": "encoders.9.layer_norm1.bias",300            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.layer_norm1.weight": "encoders.9.layer_norm1.weight",301            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.layer_norm2.bias": "encoders.9.layer_norm2.bias",302            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.layer_norm2.weight": "encoders.9.layer_norm2.weight",303            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.mlp.fc1.bias": "encoders.9.fc1.bias",304            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.mlp.fc1.weight": "encoders.9.fc1.weight",305            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.mlp.fc2.bias": "encoders.9.fc2.bias",306            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.mlp.fc2.weight": "encoders.9.fc2.weight",307            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.k_proj.bias": "encoders.9.attn.to_k.bias",308            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.k_proj.weight": "encoders.9.attn.to_k.weight",309            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.out_proj.bias": "encoders.9.attn.to_out.bias",310            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.out_proj.weight": "encoders.9.attn.to_out.weight",311            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.q_proj.bias": "encoders.9.attn.to_q.bias",312            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.q_proj.weight": "encoders.9.attn.to_q.weight",313            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.v_proj.bias": "encoders.9.attn.to_v.bias",314            "text_encoders.clip_l.transformer.text_model.encoder.layers.9.self_attn.v_proj.weight": "encoders.9.attn.to_v.weight",315            "text_encoders.clip_l.transformer.text_model.final_layer_norm.bias": "final_layer_norm.bias",316            "text_encoders.clip_l.transformer.text_model.final_layer_norm.weight": "final_layer_norm.weight",317        }318        state_dict_ = {}319        for name in state_dict:320            if name in rename_dict:321                param = state_dict[name]322                if name == "text_encoders.clip_l.transformer.text_model.embeddings.position_embedding.weight":323                    param = param.reshape((1, param.shape[0], param.shape[1]))324                state_dict_[rename_dict[name]] = param325        return state_dict_326    327 328 329class SD3TextEncoder2StateDictConverter(SDXLTextEncoder2StateDictConverter):330    def __init__(self):331        pass332 333    def from_diffusers(self, state_dict):334        return super().from_diffusers(state_dict)335    336    def from_civitai(self, state_dict):337        rename_dict = {338            "text_encoders.clip_g.transformer.text_model.embeddings.position_embedding.weight": "position_embeds",339            "text_encoders.clip_g.transformer.text_model.embeddings.token_embedding.weight": "token_embedding.weight",340            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.layer_norm1.bias": "encoders.0.layer_norm1.bias",341            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.layer_norm1.weight": "encoders.0.layer_norm1.weight",342            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.layer_norm2.bias": "encoders.0.layer_norm2.bias",343            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.layer_norm2.weight": "encoders.0.layer_norm2.weight",344            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.mlp.fc1.bias": "encoders.0.fc1.bias",345            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.mlp.fc1.weight": "encoders.0.fc1.weight",346            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.mlp.fc2.bias": "encoders.0.fc2.bias",347            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.mlp.fc2.weight": "encoders.0.fc2.weight",348            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.k_proj.bias": "encoders.0.attn.to_k.bias",349            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.k_proj.weight": "encoders.0.attn.to_k.weight",350            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.out_proj.bias": "encoders.0.attn.to_out.bias",351            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.out_proj.weight": "encoders.0.attn.to_out.weight",352            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.q_proj.bias": "encoders.0.attn.to_q.bias",353            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.q_proj.weight": "encoders.0.attn.to_q.weight",354            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.v_proj.bias": "encoders.0.attn.to_v.bias",355            "text_encoders.clip_g.transformer.text_model.encoder.layers.0.self_attn.v_proj.weight": "encoders.0.attn.to_v.weight",356            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.layer_norm1.bias": "encoders.1.layer_norm1.bias",357            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.layer_norm1.weight": "encoders.1.layer_norm1.weight",358            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.layer_norm2.bias": "encoders.1.layer_norm2.bias",359            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.layer_norm2.weight": "encoders.1.layer_norm2.weight",360            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.mlp.fc1.bias": "encoders.1.fc1.bias",361            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.mlp.fc1.weight": "encoders.1.fc1.weight",362            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.mlp.fc2.bias": "encoders.1.fc2.bias",363            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.mlp.fc2.weight": "encoders.1.fc2.weight",364            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.k_proj.bias": "encoders.1.attn.to_k.bias",365            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.k_proj.weight": "encoders.1.attn.to_k.weight",366            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.out_proj.bias": "encoders.1.attn.to_out.bias",367            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.out_proj.weight": "encoders.1.attn.to_out.weight",368            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.q_proj.bias": "encoders.1.attn.to_q.bias",369            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.q_proj.weight": "encoders.1.attn.to_q.weight",370            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.v_proj.bias": "encoders.1.attn.to_v.bias",371            "text_encoders.clip_g.transformer.text_model.encoder.layers.1.self_attn.v_proj.weight": "encoders.1.attn.to_v.weight",372            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.layer_norm1.bias": "encoders.10.layer_norm1.bias",373            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.layer_norm1.weight": "encoders.10.layer_norm1.weight",374            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.layer_norm2.bias": "encoders.10.layer_norm2.bias",375            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.layer_norm2.weight": "encoders.10.layer_norm2.weight",376            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.mlp.fc1.bias": "encoders.10.fc1.bias",377            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.mlp.fc1.weight": "encoders.10.fc1.weight",378            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.mlp.fc2.bias": "encoders.10.fc2.bias",379            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.mlp.fc2.weight": "encoders.10.fc2.weight",380            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.k_proj.bias": "encoders.10.attn.to_k.bias",381            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.k_proj.weight": "encoders.10.attn.to_k.weight",382            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.out_proj.bias": "encoders.10.attn.to_out.bias",383            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.out_proj.weight": "encoders.10.attn.to_out.weight",384            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.q_proj.bias": "encoders.10.attn.to_q.bias",385            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.q_proj.weight": "encoders.10.attn.to_q.weight",386            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.v_proj.bias": "encoders.10.attn.to_v.bias",387            "text_encoders.clip_g.transformer.text_model.encoder.layers.10.self_attn.v_proj.weight": "encoders.10.attn.to_v.weight",388            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.layer_norm1.bias": "encoders.11.layer_norm1.bias",389            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.layer_norm1.weight": "encoders.11.layer_norm1.weight",390            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.layer_norm2.bias": "encoders.11.layer_norm2.bias",391            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.layer_norm2.weight": "encoders.11.layer_norm2.weight",392            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.mlp.fc1.bias": "encoders.11.fc1.bias",393            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.mlp.fc1.weight": "encoders.11.fc1.weight",394            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.mlp.fc2.bias": "encoders.11.fc2.bias",395            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.mlp.fc2.weight": "encoders.11.fc2.weight",396            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.k_proj.bias": "encoders.11.attn.to_k.bias",397            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.k_proj.weight": "encoders.11.attn.to_k.weight",398            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.out_proj.bias": "encoders.11.attn.to_out.bias",399            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.out_proj.weight": "encoders.11.attn.to_out.weight",400            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.q_proj.bias": "encoders.11.attn.to_q.bias",401            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.q_proj.weight": "encoders.11.attn.to_q.weight",402            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.v_proj.bias": "encoders.11.attn.to_v.bias",403            "text_encoders.clip_g.transformer.text_model.encoder.layers.11.self_attn.v_proj.weight": "encoders.11.attn.to_v.weight",404            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.layer_norm1.bias": "encoders.12.layer_norm1.bias",405            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.layer_norm1.weight": "encoders.12.layer_norm1.weight",406            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.layer_norm2.bias": "encoders.12.layer_norm2.bias",407            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.layer_norm2.weight": "encoders.12.layer_norm2.weight",408            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.mlp.fc1.bias": "encoders.12.fc1.bias",409            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.mlp.fc1.weight": "encoders.12.fc1.weight",410            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.mlp.fc2.bias": "encoders.12.fc2.bias",411            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.mlp.fc2.weight": "encoders.12.fc2.weight",412            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.k_proj.bias": "encoders.12.attn.to_k.bias",413            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.k_proj.weight": "encoders.12.attn.to_k.weight",414            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.out_proj.bias": "encoders.12.attn.to_out.bias",415            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.out_proj.weight": "encoders.12.attn.to_out.weight",416            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.q_proj.bias": "encoders.12.attn.to_q.bias",417            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.q_proj.weight": "encoders.12.attn.to_q.weight",418            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.v_proj.bias": "encoders.12.attn.to_v.bias",419            "text_encoders.clip_g.transformer.text_model.encoder.layers.12.self_attn.v_proj.weight": "encoders.12.attn.to_v.weight",420            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.layer_norm1.bias": "encoders.13.layer_norm1.bias",421            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.layer_norm1.weight": "encoders.13.layer_norm1.weight",422            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.layer_norm2.bias": "encoders.13.layer_norm2.bias",423            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.layer_norm2.weight": "encoders.13.layer_norm2.weight",424            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.mlp.fc1.bias": "encoders.13.fc1.bias",425            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.mlp.fc1.weight": "encoders.13.fc1.weight",426            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.mlp.fc2.bias": "encoders.13.fc2.bias",427            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.mlp.fc2.weight": "encoders.13.fc2.weight",428            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.k_proj.bias": "encoders.13.attn.to_k.bias",429            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.k_proj.weight": "encoders.13.attn.to_k.weight",430            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.out_proj.bias": "encoders.13.attn.to_out.bias",431            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.out_proj.weight": "encoders.13.attn.to_out.weight",432            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.q_proj.bias": "encoders.13.attn.to_q.bias",433            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.q_proj.weight": "encoders.13.attn.to_q.weight",434            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.v_proj.bias": "encoders.13.attn.to_v.bias",435            "text_encoders.clip_g.transformer.text_model.encoder.layers.13.self_attn.v_proj.weight": "encoders.13.attn.to_v.weight",436            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.layer_norm1.bias": "encoders.14.layer_norm1.bias",437            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.layer_norm1.weight": "encoders.14.layer_norm1.weight",438            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.layer_norm2.bias": "encoders.14.layer_norm2.bias",439            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.layer_norm2.weight": "encoders.14.layer_norm2.weight",440            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.mlp.fc1.bias": "encoders.14.fc1.bias",441            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.mlp.fc1.weight": "encoders.14.fc1.weight",442            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.mlp.fc2.bias": "encoders.14.fc2.bias",443            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.mlp.fc2.weight": "encoders.14.fc2.weight",444            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.k_proj.bias": "encoders.14.attn.to_k.bias",445            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.k_proj.weight": "encoders.14.attn.to_k.weight",446            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.out_proj.bias": "encoders.14.attn.to_out.bias",447            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.out_proj.weight": "encoders.14.attn.to_out.weight",448            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.q_proj.bias": "encoders.14.attn.to_q.bias",449            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.q_proj.weight": "encoders.14.attn.to_q.weight",450            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.v_proj.bias": "encoders.14.attn.to_v.bias",451            "text_encoders.clip_g.transformer.text_model.encoder.layers.14.self_attn.v_proj.weight": "encoders.14.attn.to_v.weight",452            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.layer_norm1.bias": "encoders.15.layer_norm1.bias",453            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.layer_norm1.weight": "encoders.15.layer_norm1.weight",454            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.layer_norm2.bias": "encoders.15.layer_norm2.bias",455            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.layer_norm2.weight": "encoders.15.layer_norm2.weight",456            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.mlp.fc1.bias": "encoders.15.fc1.bias",457            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.mlp.fc1.weight": "encoders.15.fc1.weight",458            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.mlp.fc2.bias": "encoders.15.fc2.bias",459            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.mlp.fc2.weight": "encoders.15.fc2.weight",460            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.k_proj.bias": "encoders.15.attn.to_k.bias",461            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.k_proj.weight": "encoders.15.attn.to_k.weight",462            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.out_proj.bias": "encoders.15.attn.to_out.bias",463            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.out_proj.weight": "encoders.15.attn.to_out.weight",464            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.q_proj.bias": "encoders.15.attn.to_q.bias",465            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.q_proj.weight": "encoders.15.attn.to_q.weight",466            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.v_proj.bias": "encoders.15.attn.to_v.bias",467            "text_encoders.clip_g.transformer.text_model.encoder.layers.15.self_attn.v_proj.weight": "encoders.15.attn.to_v.weight",468            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.layer_norm1.bias": "encoders.16.layer_norm1.bias",469            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.layer_norm1.weight": "encoders.16.layer_norm1.weight",470            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.layer_norm2.bias": "encoders.16.layer_norm2.bias",471            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.layer_norm2.weight": "encoders.16.layer_norm2.weight",472            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.mlp.fc1.bias": "encoders.16.fc1.bias",473            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.mlp.fc1.weight": "encoders.16.fc1.weight",474            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.mlp.fc2.bias": "encoders.16.fc2.bias",475            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.mlp.fc2.weight": "encoders.16.fc2.weight",476            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.k_proj.bias": "encoders.16.attn.to_k.bias",477            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.k_proj.weight": "encoders.16.attn.to_k.weight",478            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.out_proj.bias": "encoders.16.attn.to_out.bias",479            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.out_proj.weight": "encoders.16.attn.to_out.weight",480            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.q_proj.bias": "encoders.16.attn.to_q.bias",481            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.q_proj.weight": "encoders.16.attn.to_q.weight",482            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.v_proj.bias": "encoders.16.attn.to_v.bias",483            "text_encoders.clip_g.transformer.text_model.encoder.layers.16.self_attn.v_proj.weight": "encoders.16.attn.to_v.weight",484            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.layer_norm1.bias": "encoders.17.layer_norm1.bias",485            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.layer_norm1.weight": "encoders.17.layer_norm1.weight",486            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.layer_norm2.bias": "encoders.17.layer_norm2.bias",487            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.layer_norm2.weight": "encoders.17.layer_norm2.weight",488            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.mlp.fc1.bias": "encoders.17.fc1.bias",489            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.mlp.fc1.weight": "encoders.17.fc1.weight",490            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.mlp.fc2.bias": "encoders.17.fc2.bias",491            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.mlp.fc2.weight": "encoders.17.fc2.weight",492            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.k_proj.bias": "encoders.17.attn.to_k.bias",493            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.k_proj.weight": "encoders.17.attn.to_k.weight",494            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.out_proj.bias": "encoders.17.attn.to_out.bias",495            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.out_proj.weight": "encoders.17.attn.to_out.weight",496            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.q_proj.bias": "encoders.17.attn.to_q.bias",497            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.q_proj.weight": "encoders.17.attn.to_q.weight",498            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.v_proj.bias": "encoders.17.attn.to_v.bias",499            "text_encoders.clip_g.transformer.text_model.encoder.layers.17.self_attn.v_proj.weight": "encoders.17.attn.to_v.weight",500            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.layer_norm1.bias": "encoders.18.layer_norm1.bias",501            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.layer_norm1.weight": "encoders.18.layer_norm1.weight",502            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.layer_norm2.bias": "encoders.18.layer_norm2.bias",503            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.layer_norm2.weight": "encoders.18.layer_norm2.weight",504            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.mlp.fc1.bias": "encoders.18.fc1.bias",505            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.mlp.fc1.weight": "encoders.18.fc1.weight",506            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.mlp.fc2.bias": "encoders.18.fc2.bias",507            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.mlp.fc2.weight": "encoders.18.fc2.weight",508            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.k_proj.bias": "encoders.18.attn.to_k.bias",509            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.k_proj.weight": "encoders.18.attn.to_k.weight",510            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.out_proj.bias": "encoders.18.attn.to_out.bias",511            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.out_proj.weight": "encoders.18.attn.to_out.weight",512            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.q_proj.bias": "encoders.18.attn.to_q.bias",513            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.q_proj.weight": "encoders.18.attn.to_q.weight",514            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.v_proj.bias": "encoders.18.attn.to_v.bias",515            "text_encoders.clip_g.transformer.text_model.encoder.layers.18.self_attn.v_proj.weight": "encoders.18.attn.to_v.weight",516            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.layer_norm1.bias": "encoders.19.layer_norm1.bias",517            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.layer_norm1.weight": "encoders.19.layer_norm1.weight",518            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.layer_norm2.bias": "encoders.19.layer_norm2.bias",519            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.layer_norm2.weight": "encoders.19.layer_norm2.weight",520            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.mlp.fc1.bias": "encoders.19.fc1.bias",521            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.mlp.fc1.weight": "encoders.19.fc1.weight",522            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.mlp.fc2.bias": "encoders.19.fc2.bias",523            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.mlp.fc2.weight": "encoders.19.fc2.weight",524            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.k_proj.bias": "encoders.19.attn.to_k.bias",525            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.k_proj.weight": "encoders.19.attn.to_k.weight",526            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.out_proj.bias": "encoders.19.attn.to_out.bias",527            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.out_proj.weight": "encoders.19.attn.to_out.weight",528            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.q_proj.bias": "encoders.19.attn.to_q.bias",529            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.q_proj.weight": "encoders.19.attn.to_q.weight",530            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.v_proj.bias": "encoders.19.attn.to_v.bias",531            "text_encoders.clip_g.transformer.text_model.encoder.layers.19.self_attn.v_proj.weight": "encoders.19.attn.to_v.weight",532            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.layer_norm1.bias": "encoders.2.layer_norm1.bias",533            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.layer_norm1.weight": "encoders.2.layer_norm1.weight",534            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.layer_norm2.bias": "encoders.2.layer_norm2.bias",535            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.layer_norm2.weight": "encoders.2.layer_norm2.weight",536            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.mlp.fc1.bias": "encoders.2.fc1.bias",537            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.mlp.fc1.weight": "encoders.2.fc1.weight",538            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.mlp.fc2.bias": "encoders.2.fc2.bias",539            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.mlp.fc2.weight": "encoders.2.fc2.weight",540            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.k_proj.bias": "encoders.2.attn.to_k.bias",541            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.k_proj.weight": "encoders.2.attn.to_k.weight",542            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.out_proj.bias": "encoders.2.attn.to_out.bias",543            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.out_proj.weight": "encoders.2.attn.to_out.weight",544            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.q_proj.bias": "encoders.2.attn.to_q.bias",545            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.q_proj.weight": "encoders.2.attn.to_q.weight",546            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.v_proj.bias": "encoders.2.attn.to_v.bias",547            "text_encoders.clip_g.transformer.text_model.encoder.layers.2.self_attn.v_proj.weight": "encoders.2.attn.to_v.weight",548            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.layer_norm1.bias": "encoders.20.layer_norm1.bias",549            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.layer_norm1.weight": "encoders.20.layer_norm1.weight",550            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.layer_norm2.bias": "encoders.20.layer_norm2.bias",551            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.layer_norm2.weight": "encoders.20.layer_norm2.weight",552            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.mlp.fc1.bias": "encoders.20.fc1.bias",553            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.mlp.fc1.weight": "encoders.20.fc1.weight",554            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.mlp.fc2.bias": "encoders.20.fc2.bias",555            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.mlp.fc2.weight": "encoders.20.fc2.weight",556            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.k_proj.bias": "encoders.20.attn.to_k.bias",557            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.k_proj.weight": "encoders.20.attn.to_k.weight",558            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.out_proj.bias": "encoders.20.attn.to_out.bias",559            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.out_proj.weight": "encoders.20.attn.to_out.weight",560            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.q_proj.bias": "encoders.20.attn.to_q.bias",561            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.q_proj.weight": "encoders.20.attn.to_q.weight",562            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.v_proj.bias": "encoders.20.attn.to_v.bias",563            "text_encoders.clip_g.transformer.text_model.encoder.layers.20.self_attn.v_proj.weight": "encoders.20.attn.to_v.weight",564            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.layer_norm1.bias": "encoders.21.layer_norm1.bias",565            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.layer_norm1.weight": "encoders.21.layer_norm1.weight",566            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.layer_norm2.bias": "encoders.21.layer_norm2.bias",567            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.layer_norm2.weight": "encoders.21.layer_norm2.weight",568            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.mlp.fc1.bias": "encoders.21.fc1.bias",569            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.mlp.fc1.weight": "encoders.21.fc1.weight",570            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.mlp.fc2.bias": "encoders.21.fc2.bias",571            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.mlp.fc2.weight": "encoders.21.fc2.weight",572            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.k_proj.bias": "encoders.21.attn.to_k.bias",573            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.k_proj.weight": "encoders.21.attn.to_k.weight",574            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.out_proj.bias": "encoders.21.attn.to_out.bias",575            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.out_proj.weight": "encoders.21.attn.to_out.weight",576            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.q_proj.bias": "encoders.21.attn.to_q.bias",577            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.q_proj.weight": "encoders.21.attn.to_q.weight",578            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.v_proj.bias": "encoders.21.attn.to_v.bias",579            "text_encoders.clip_g.transformer.text_model.encoder.layers.21.self_attn.v_proj.weight": "encoders.21.attn.to_v.weight",580            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.layer_norm1.bias": "encoders.22.layer_norm1.bias",581            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.layer_norm1.weight": "encoders.22.layer_norm1.weight",582            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.layer_norm2.bias": "encoders.22.layer_norm2.bias",583            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.layer_norm2.weight": "encoders.22.layer_norm2.weight",584            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.mlp.fc1.bias": "encoders.22.fc1.bias",585            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.mlp.fc1.weight": "encoders.22.fc1.weight",586            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.mlp.fc2.bias": "encoders.22.fc2.bias",587            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.mlp.fc2.weight": "encoders.22.fc2.weight",588            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.k_proj.bias": "encoders.22.attn.to_k.bias",589            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.k_proj.weight": "encoders.22.attn.to_k.weight",590            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.out_proj.bias": "encoders.22.attn.to_out.bias",591            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.out_proj.weight": "encoders.22.attn.to_out.weight",592            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.q_proj.bias": "encoders.22.attn.to_q.bias",593            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.q_proj.weight": "encoders.22.attn.to_q.weight",594            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.v_proj.bias": "encoders.22.attn.to_v.bias",595            "text_encoders.clip_g.transformer.text_model.encoder.layers.22.self_attn.v_proj.weight": "encoders.22.attn.to_v.weight",596            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.layer_norm1.bias": "encoders.23.layer_norm1.bias",597            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.layer_norm1.weight": "encoders.23.layer_norm1.weight",598            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.layer_norm2.bias": "encoders.23.layer_norm2.bias",599            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.layer_norm2.weight": "encoders.23.layer_norm2.weight",600            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.mlp.fc1.bias": "encoders.23.fc1.bias",601            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.mlp.fc1.weight": "encoders.23.fc1.weight",602            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.mlp.fc2.bias": "encoders.23.fc2.bias",603            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.mlp.fc2.weight": "encoders.23.fc2.weight",604            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.k_proj.bias": "encoders.23.attn.to_k.bias",605            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.k_proj.weight": "encoders.23.attn.to_k.weight",606            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.out_proj.bias": "encoders.23.attn.to_out.bias",607            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.out_proj.weight": "encoders.23.attn.to_out.weight",608            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.q_proj.bias": "encoders.23.attn.to_q.bias",609            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.q_proj.weight": "encoders.23.attn.to_q.weight",610            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.v_proj.bias": "encoders.23.attn.to_v.bias",611            "text_encoders.clip_g.transformer.text_model.encoder.layers.23.self_attn.v_proj.weight": "encoders.23.attn.to_v.weight",612            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.layer_norm1.bias": "encoders.24.layer_norm1.bias",613            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.layer_norm1.weight": "encoders.24.layer_norm1.weight",614            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.layer_norm2.bias": "encoders.24.layer_norm2.bias",615            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.layer_norm2.weight": "encoders.24.layer_norm2.weight",616            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.mlp.fc1.bias": "encoders.24.fc1.bias",617            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.mlp.fc1.weight": "encoders.24.fc1.weight",618            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.mlp.fc2.bias": "encoders.24.fc2.bias",619            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.mlp.fc2.weight": "encoders.24.fc2.weight",620            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.k_proj.bias": "encoders.24.attn.to_k.bias",621            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.k_proj.weight": "encoders.24.attn.to_k.weight",622            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.out_proj.bias": "encoders.24.attn.to_out.bias",623            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.out_proj.weight": "encoders.24.attn.to_out.weight",624            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.q_proj.bias": "encoders.24.attn.to_q.bias",625            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.q_proj.weight": "encoders.24.attn.to_q.weight",626            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.v_proj.bias": "encoders.24.attn.to_v.bias",627            "text_encoders.clip_g.transformer.text_model.encoder.layers.24.self_attn.v_proj.weight": "encoders.24.attn.to_v.weight",628            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.layer_norm1.bias": "encoders.25.layer_norm1.bias",629            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.layer_norm1.weight": "encoders.25.layer_norm1.weight",630            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.layer_norm2.bias": "encoders.25.layer_norm2.bias",631            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.layer_norm2.weight": "encoders.25.layer_norm2.weight",632            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.mlp.fc1.bias": "encoders.25.fc1.bias",633            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.mlp.fc1.weight": "encoders.25.fc1.weight",634            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.mlp.fc2.bias": "encoders.25.fc2.bias",635            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.mlp.fc2.weight": "encoders.25.fc2.weight",636            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.k_proj.bias": "encoders.25.attn.to_k.bias",637            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.k_proj.weight": "encoders.25.attn.to_k.weight",638            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.out_proj.bias": "encoders.25.attn.to_out.bias",639            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.out_proj.weight": "encoders.25.attn.to_out.weight",640            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.q_proj.bias": "encoders.25.attn.to_q.bias",641            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.q_proj.weight": "encoders.25.attn.to_q.weight",642            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.v_proj.bias": "encoders.25.attn.to_v.bias",643            "text_encoders.clip_g.transformer.text_model.encoder.layers.25.self_attn.v_proj.weight": "encoders.25.attn.to_v.weight",644            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.layer_norm1.bias": "encoders.26.layer_norm1.bias",645            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.layer_norm1.weight": "encoders.26.layer_norm1.weight",646            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.layer_norm2.bias": "encoders.26.layer_norm2.bias",647            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.layer_norm2.weight": "encoders.26.layer_norm2.weight",648            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.mlp.fc1.bias": "encoders.26.fc1.bias",649            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.mlp.fc1.weight": "encoders.26.fc1.weight",650            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.mlp.fc2.bias": "encoders.26.fc2.bias",651            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.mlp.fc2.weight": "encoders.26.fc2.weight",652            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.k_proj.bias": "encoders.26.attn.to_k.bias",653            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.k_proj.weight": "encoders.26.attn.to_k.weight",654            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.out_proj.bias": "encoders.26.attn.to_out.bias",655            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.out_proj.weight": "encoders.26.attn.to_out.weight",656            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.q_proj.bias": "encoders.26.attn.to_q.bias",657            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.q_proj.weight": "encoders.26.attn.to_q.weight",658            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.v_proj.bias": "encoders.26.attn.to_v.bias",659            "text_encoders.clip_g.transformer.text_model.encoder.layers.26.self_attn.v_proj.weight": "encoders.26.attn.to_v.weight",660            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.layer_norm1.bias": "encoders.27.layer_norm1.bias",661            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.layer_norm1.weight": "encoders.27.layer_norm1.weight",662            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.layer_norm2.bias": "encoders.27.layer_norm2.bias",663            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.layer_norm2.weight": "encoders.27.layer_norm2.weight",664            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.mlp.fc1.bias": "encoders.27.fc1.bias",665            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.mlp.fc1.weight": "encoders.27.fc1.weight",666            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.mlp.fc2.bias": "encoders.27.fc2.bias",667            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.mlp.fc2.weight": "encoders.27.fc2.weight",668            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.k_proj.bias": "encoders.27.attn.to_k.bias",669            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.k_proj.weight": "encoders.27.attn.to_k.weight",670            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.out_proj.bias": "encoders.27.attn.to_out.bias",671            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.out_proj.weight": "encoders.27.attn.to_out.weight",672            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.q_proj.bias": "encoders.27.attn.to_q.bias",673            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.q_proj.weight": "encoders.27.attn.to_q.weight",674            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.v_proj.bias": "encoders.27.attn.to_v.bias",675            "text_encoders.clip_g.transformer.text_model.encoder.layers.27.self_attn.v_proj.weight": "encoders.27.attn.to_v.weight",676            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.layer_norm1.bias": "encoders.28.layer_norm1.bias",677            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.layer_norm1.weight": "encoders.28.layer_norm1.weight",678            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.layer_norm2.bias": "encoders.28.layer_norm2.bias",679            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.layer_norm2.weight": "encoders.28.layer_norm2.weight",680            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.mlp.fc1.bias": "encoders.28.fc1.bias",681            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.mlp.fc1.weight": "encoders.28.fc1.weight",682            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.mlp.fc2.bias": "encoders.28.fc2.bias",683            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.mlp.fc2.weight": "encoders.28.fc2.weight",684            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.k_proj.bias": "encoders.28.attn.to_k.bias",685            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.k_proj.weight": "encoders.28.attn.to_k.weight",686            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.out_proj.bias": "encoders.28.attn.to_out.bias",687            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.out_proj.weight": "encoders.28.attn.to_out.weight",688            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.q_proj.bias": "encoders.28.attn.to_q.bias",689            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.q_proj.weight": "encoders.28.attn.to_q.weight",690            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.v_proj.bias": "encoders.28.attn.to_v.bias",691            "text_encoders.clip_g.transformer.text_model.encoder.layers.28.self_attn.v_proj.weight": "encoders.28.attn.to_v.weight",692            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.layer_norm1.bias": "encoders.29.layer_norm1.bias",693            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.layer_norm1.weight": "encoders.29.layer_norm1.weight",694            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.layer_norm2.bias": "encoders.29.layer_norm2.bias",695            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.layer_norm2.weight": "encoders.29.layer_norm2.weight",696            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.mlp.fc1.bias": "encoders.29.fc1.bias",697            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.mlp.fc1.weight": "encoders.29.fc1.weight",698            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.mlp.fc2.bias": "encoders.29.fc2.bias",699            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.mlp.fc2.weight": "encoders.29.fc2.weight",700            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.k_proj.bias": "encoders.29.attn.to_k.bias",701            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.k_proj.weight": "encoders.29.attn.to_k.weight",702            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.out_proj.bias": "encoders.29.attn.to_out.bias",703            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.out_proj.weight": "encoders.29.attn.to_out.weight",704            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.q_proj.bias": "encoders.29.attn.to_q.bias",705            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.q_proj.weight": "encoders.29.attn.to_q.weight",706            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.v_proj.bias": "encoders.29.attn.to_v.bias",707            "text_encoders.clip_g.transformer.text_model.encoder.layers.29.self_attn.v_proj.weight": "encoders.29.attn.to_v.weight",708            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.layer_norm1.bias": "encoders.3.layer_norm1.bias",709            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.layer_norm1.weight": "encoders.3.layer_norm1.weight",710            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.layer_norm2.bias": "encoders.3.layer_norm2.bias",711            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.layer_norm2.weight": "encoders.3.layer_norm2.weight",712            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.mlp.fc1.bias": "encoders.3.fc1.bias",713            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.mlp.fc1.weight": "encoders.3.fc1.weight",714            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.mlp.fc2.bias": "encoders.3.fc2.bias",715            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.mlp.fc2.weight": "encoders.3.fc2.weight",716            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.k_proj.bias": "encoders.3.attn.to_k.bias",717            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.k_proj.weight": "encoders.3.attn.to_k.weight",718            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.out_proj.bias": "encoders.3.attn.to_out.bias",719            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.out_proj.weight": "encoders.3.attn.to_out.weight",720            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.q_proj.bias": "encoders.3.attn.to_q.bias",721            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.q_proj.weight": "encoders.3.attn.to_q.weight",722            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.v_proj.bias": "encoders.3.attn.to_v.bias",723            "text_encoders.clip_g.transformer.text_model.encoder.layers.3.self_attn.v_proj.weight": "encoders.3.attn.to_v.weight",724            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.layer_norm1.bias": "encoders.30.layer_norm1.bias",725            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.layer_norm1.weight": "encoders.30.layer_norm1.weight",726            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.layer_norm2.bias": "encoders.30.layer_norm2.bias",727            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.layer_norm2.weight": "encoders.30.layer_norm2.weight",728            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.mlp.fc1.bias": "encoders.30.fc1.bias",729            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.mlp.fc1.weight": "encoders.30.fc1.weight",730            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.mlp.fc2.bias": "encoders.30.fc2.bias",731            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.mlp.fc2.weight": "encoders.30.fc2.weight",732            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.k_proj.bias": "encoders.30.attn.to_k.bias",733            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.k_proj.weight": "encoders.30.attn.to_k.weight",734            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.out_proj.bias": "encoders.30.attn.to_out.bias",735            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.out_proj.weight": "encoders.30.attn.to_out.weight",736            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.q_proj.bias": "encoders.30.attn.to_q.bias",737            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.q_proj.weight": "encoders.30.attn.to_q.weight",738            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.v_proj.bias": "encoders.30.attn.to_v.bias",739            "text_encoders.clip_g.transformer.text_model.encoder.layers.30.self_attn.v_proj.weight": "encoders.30.attn.to_v.weight",740            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.layer_norm1.bias": "encoders.31.layer_norm1.bias",741            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.layer_norm1.weight": "encoders.31.layer_norm1.weight",742            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.layer_norm2.bias": "encoders.31.layer_norm2.bias",743            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.layer_norm2.weight": "encoders.31.layer_norm2.weight",744            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.mlp.fc1.bias": "encoders.31.fc1.bias",745            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.mlp.fc1.weight": "encoders.31.fc1.weight",746            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.mlp.fc2.bias": "encoders.31.fc2.bias",747            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.mlp.fc2.weight": "encoders.31.fc2.weight",748            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.k_proj.bias": "encoders.31.attn.to_k.bias",749            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.k_proj.weight": "encoders.31.attn.to_k.weight",750            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.out_proj.bias": "encoders.31.attn.to_out.bias",751            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.out_proj.weight": "encoders.31.attn.to_out.weight",752            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.q_proj.bias": "encoders.31.attn.to_q.bias",753            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.q_proj.weight": "encoders.31.attn.to_q.weight",754            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.v_proj.bias": "encoders.31.attn.to_v.bias",755            "text_encoders.clip_g.transformer.text_model.encoder.layers.31.self_attn.v_proj.weight": "encoders.31.attn.to_v.weight",756            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.layer_norm1.bias": "encoders.4.layer_norm1.bias",757            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.layer_norm1.weight": "encoders.4.layer_norm1.weight",758            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.layer_norm2.bias": "encoders.4.layer_norm2.bias",759            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.layer_norm2.weight": "encoders.4.layer_norm2.weight",760            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.mlp.fc1.bias": "encoders.4.fc1.bias",761            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.mlp.fc1.weight": "encoders.4.fc1.weight",762            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.mlp.fc2.bias": "encoders.4.fc2.bias",763            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.mlp.fc2.weight": "encoders.4.fc2.weight",764            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.k_proj.bias": "encoders.4.attn.to_k.bias",765            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.k_proj.weight": "encoders.4.attn.to_k.weight",766            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.out_proj.bias": "encoders.4.attn.to_out.bias",767            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.out_proj.weight": "encoders.4.attn.to_out.weight",768            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.q_proj.bias": "encoders.4.attn.to_q.bias",769            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.q_proj.weight": "encoders.4.attn.to_q.weight",770            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.v_proj.bias": "encoders.4.attn.to_v.bias",771            "text_encoders.clip_g.transformer.text_model.encoder.layers.4.self_attn.v_proj.weight": "encoders.4.attn.to_v.weight",772            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.layer_norm1.bias": "encoders.5.layer_norm1.bias",773            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.layer_norm1.weight": "encoders.5.layer_norm1.weight",774            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.layer_norm2.bias": "encoders.5.layer_norm2.bias",775            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.layer_norm2.weight": "encoders.5.layer_norm2.weight",776            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.mlp.fc1.bias": "encoders.5.fc1.bias",777            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.mlp.fc1.weight": "encoders.5.fc1.weight",778            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.mlp.fc2.bias": "encoders.5.fc2.bias",779            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.mlp.fc2.weight": "encoders.5.fc2.weight",780            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.k_proj.bias": "encoders.5.attn.to_k.bias",781            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.k_proj.weight": "encoders.5.attn.to_k.weight",782            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.out_proj.bias": "encoders.5.attn.to_out.bias",783            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.out_proj.weight": "encoders.5.attn.to_out.weight",784            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.q_proj.bias": "encoders.5.attn.to_q.bias",785            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.q_proj.weight": "encoders.5.attn.to_q.weight",786            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.v_proj.bias": "encoders.5.attn.to_v.bias",787            "text_encoders.clip_g.transformer.text_model.encoder.layers.5.self_attn.v_proj.weight": "encoders.5.attn.to_v.weight",788            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.layer_norm1.bias": "encoders.6.layer_norm1.bias",789            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.layer_norm1.weight": "encoders.6.layer_norm1.weight",790            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.layer_norm2.bias": "encoders.6.layer_norm2.bias",791            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.layer_norm2.weight": "encoders.6.layer_norm2.weight",792            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.mlp.fc1.bias": "encoders.6.fc1.bias",793            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.mlp.fc1.weight": "encoders.6.fc1.weight",794            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.mlp.fc2.bias": "encoders.6.fc2.bias",795            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.mlp.fc2.weight": "encoders.6.fc2.weight",796            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.k_proj.bias": "encoders.6.attn.to_k.bias",797            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.k_proj.weight": "encoders.6.attn.to_k.weight",798            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.out_proj.bias": "encoders.6.attn.to_out.bias",799            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.out_proj.weight": "encoders.6.attn.to_out.weight",800            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.q_proj.bias": "encoders.6.attn.to_q.bias",801            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.q_proj.weight": "encoders.6.attn.to_q.weight",802            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.v_proj.bias": "encoders.6.attn.to_v.bias",803            "text_encoders.clip_g.transformer.text_model.encoder.layers.6.self_attn.v_proj.weight": "encoders.6.attn.to_v.weight",804            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.layer_norm1.bias": "encoders.7.layer_norm1.bias",805            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.layer_norm1.weight": "encoders.7.layer_norm1.weight",806            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.layer_norm2.bias": "encoders.7.layer_norm2.bias",807            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.layer_norm2.weight": "encoders.7.layer_norm2.weight",808            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.mlp.fc1.bias": "encoders.7.fc1.bias",809            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.mlp.fc1.weight": "encoders.7.fc1.weight",810            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.mlp.fc2.bias": "encoders.7.fc2.bias",811            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.mlp.fc2.weight": "encoders.7.fc2.weight",812            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.k_proj.bias": "encoders.7.attn.to_k.bias",813            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.k_proj.weight": "encoders.7.attn.to_k.weight",814            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.out_proj.bias": "encoders.7.attn.to_out.bias",815            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.out_proj.weight": "encoders.7.attn.to_out.weight",816            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.q_proj.bias": "encoders.7.attn.to_q.bias",817            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.q_proj.weight": "encoders.7.attn.to_q.weight",818            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.v_proj.bias": "encoders.7.attn.to_v.bias",819            "text_encoders.clip_g.transformer.text_model.encoder.layers.7.self_attn.v_proj.weight": "encoders.7.attn.to_v.weight",820            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.layer_norm1.bias": "encoders.8.layer_norm1.bias",821            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.layer_norm1.weight": "encoders.8.layer_norm1.weight",822            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.layer_norm2.bias": "encoders.8.layer_norm2.bias",823            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.layer_norm2.weight": "encoders.8.layer_norm2.weight",824            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.mlp.fc1.bias": "encoders.8.fc1.bias",825            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.mlp.fc1.weight": "encoders.8.fc1.weight",826            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.mlp.fc2.bias": "encoders.8.fc2.bias",827            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.mlp.fc2.weight": "encoders.8.fc2.weight",828            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.k_proj.bias": "encoders.8.attn.to_k.bias",829            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.k_proj.weight": "encoders.8.attn.to_k.weight",830            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.out_proj.bias": "encoders.8.attn.to_out.bias",831            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.out_proj.weight": "encoders.8.attn.to_out.weight",832            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.q_proj.bias": "encoders.8.attn.to_q.bias",833            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.q_proj.weight": "encoders.8.attn.to_q.weight",834            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.v_proj.bias": "encoders.8.attn.to_v.bias",835            "text_encoders.clip_g.transformer.text_model.encoder.layers.8.self_attn.v_proj.weight": "encoders.8.attn.to_v.weight",836            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.layer_norm1.bias": "encoders.9.layer_norm1.bias",837            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.layer_norm1.weight": "encoders.9.layer_norm1.weight",838            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.layer_norm2.bias": "encoders.9.layer_norm2.bias",839            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.layer_norm2.weight": "encoders.9.layer_norm2.weight",840            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.mlp.fc1.bias": "encoders.9.fc1.bias",841            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.mlp.fc1.weight": "encoders.9.fc1.weight",842            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.mlp.fc2.bias": "encoders.9.fc2.bias",843            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.mlp.fc2.weight": "encoders.9.fc2.weight",844            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.k_proj.bias": "encoders.9.attn.to_k.bias",845            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.k_proj.weight": "encoders.9.attn.to_k.weight",846            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.out_proj.bias": "encoders.9.attn.to_out.bias",847            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.out_proj.weight": "encoders.9.attn.to_out.weight",848            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.q_proj.bias": "encoders.9.attn.to_q.bias",849            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.q_proj.weight": "encoders.9.attn.to_q.weight",850            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.v_proj.bias": "encoders.9.attn.to_v.bias",851            "text_encoders.clip_g.transformer.text_model.encoder.layers.9.self_attn.v_proj.weight": "encoders.9.attn.to_v.weight",852            "text_encoders.clip_g.transformer.text_model.final_layer_norm.bias": "final_layer_norm.bias",853            "text_encoders.clip_g.transformer.text_model.final_layer_norm.weight": "final_layer_norm.weight",854            "text_encoders.clip_g.transformer.text_projection.weight": "text_projection.weight",855        }856        state_dict_ = {}857        for name in state_dict:858            if name in rename_dict:859                param = state_dict[name]860                if name == "text_encoders.clip_g.transformer.text_model.embeddings.position_embedding.weight":861                    param = param.reshape((1, param.shape[0], param.shape[1]))862                state_dict_[rename_dict[name]] = param863        return state_dict_864 865 866 867class SD3TextEncoder3StateDictConverter():868    def __init__(self):869        pass870 871    def from_diffusers(self, state_dict):872        state_dict_ = state_dict873        return state_dict_874    875    def from_civitai(self, state_dict):876        prefix = "text_encoders.t5xxl.transformer."877        state_dict_ = {name[len(prefix):]: param for name, param in state_dict.items() if name.startswith(prefix)}878        if len(state_dict_) > 0:879            return self.from_diffusers(state_dict_)880        name_list = [881            "encoder.block.0.layer.0.SelfAttention.k.weight",882            "encoder.block.0.layer.0.SelfAttention.o.weight",883            "encoder.block.0.layer.0.SelfAttention.q.weight",884            "encoder.block.0.layer.0.SelfAttention.relative_attention_bias.weight",885            "encoder.block.0.layer.0.SelfAttention.v.weight",886            "encoder.block.0.layer.0.layer_norm.weight",887            "encoder.block.0.layer.1.DenseReluDense.wi_0.weight",888            "encoder.block.0.layer.1.DenseReluDense.wi_1.weight",889            "encoder.block.0.layer.1.DenseReluDense.wo.weight",890            "encoder.block.0.layer.1.layer_norm.weight",891            "encoder.block.1.layer.0.SelfAttention.k.weight",892            "encoder.block.1.layer.0.SelfAttention.o.weight",893            "encoder.block.1.layer.0.SelfAttention.q.weight",894            "encoder.block.1.layer.0.SelfAttention.v.weight",895            "encoder.block.1.layer.0.layer_norm.weight",896            "encoder.block.1.layer.1.DenseReluDense.wi_0.weight",897            "encoder.block.1.layer.1.DenseReluDense.wi_1.weight",898            "encoder.block.1.layer.1.DenseReluDense.wo.weight",899            "encoder.block.1.layer.1.layer_norm.weight",900            "encoder.block.10.layer.0.SelfAttention.k.weight",901            "encoder.block.10.layer.0.SelfAttention.o.weight",902            "encoder.block.10.layer.0.SelfAttention.q.weight",903            "encoder.block.10.layer.0.SelfAttention.v.weight",904            "encoder.block.10.layer.0.layer_norm.weight",905            "encoder.block.10.layer.1.DenseReluDense.wi_0.weight",906            "encoder.block.10.layer.1.DenseReluDense.wi_1.weight",907            "encoder.block.10.layer.1.DenseReluDense.wo.weight",908            "encoder.block.10.layer.1.layer_norm.weight",909            "encoder.block.11.layer.0.SelfAttention.k.weight",910            "encoder.block.11.layer.0.SelfAttention.o.weight",911            "encoder.block.11.layer.0.SelfAttention.q.weight",912            "encoder.block.11.layer.0.SelfAttention.v.weight",913            "encoder.block.11.layer.0.layer_norm.weight",914            "encoder.block.11.layer.1.DenseReluDense.wi_0.weight",915            "encoder.block.11.layer.1.DenseReluDense.wi_1.weight",916            "encoder.block.11.layer.1.DenseReluDense.wo.weight",917            "encoder.block.11.layer.1.layer_norm.weight",918            "encoder.block.12.layer.0.SelfAttention.k.weight",919            "encoder.block.12.layer.0.SelfAttention.o.weight",920            "encoder.block.12.layer.0.SelfAttention.q.weight",921            "encoder.block.12.layer.0.SelfAttention.v.weight",922            "encoder.block.12.layer.0.layer_norm.weight",923            "encoder.block.12.layer.1.DenseReluDense.wi_0.weight",924            "encoder.block.12.layer.1.DenseReluDense.wi_1.weight",925            "encoder.block.12.layer.1.DenseReluDense.wo.weight",926            "encoder.block.12.layer.1.layer_norm.weight",927            "encoder.block.13.layer.0.SelfAttention.k.weight",928            "encoder.block.13.layer.0.SelfAttention.o.weight",929            "encoder.block.13.layer.0.SelfAttention.q.weight",930            "encoder.block.13.layer.0.SelfAttention.v.weight",931            "encoder.block.13.layer.0.layer_norm.weight",932            "encoder.block.13.layer.1.DenseReluDense.wi_0.weight",933            "encoder.block.13.layer.1.DenseReluDense.wi_1.weight",934            "encoder.block.13.layer.1.DenseReluDense.wo.weight",935            "encoder.block.13.layer.1.layer_norm.weight",936            "encoder.block.14.layer.0.SelfAttention.k.weight",937            "encoder.block.14.layer.0.SelfAttention.o.weight",938            "encoder.block.14.layer.0.SelfAttention.q.weight",939            "encoder.block.14.layer.0.SelfAttention.v.weight",940            "encoder.block.14.layer.0.layer_norm.weight",941            "encoder.block.14.layer.1.DenseReluDense.wi_0.weight",942            "encoder.block.14.layer.1.DenseReluDense.wi_1.weight",943            "encoder.block.14.layer.1.DenseReluDense.wo.weight",944            "encoder.block.14.layer.1.layer_norm.weight",945            "encoder.block.15.layer.0.SelfAttention.k.weight",946            "encoder.block.15.layer.0.SelfAttention.o.weight",947            "encoder.block.15.layer.0.SelfAttention.q.weight",948            "encoder.block.15.layer.0.SelfAttention.v.weight",949            "encoder.block.15.layer.0.layer_norm.weight",950            "encoder.block.15.layer.1.DenseReluDense.wi_0.weight",951            "encoder.block.15.layer.1.DenseReluDense.wi_1.weight",952            "encoder.block.15.layer.1.DenseReluDense.wo.weight",953            "encoder.block.15.layer.1.layer_norm.weight",954            "encoder.block.16.layer.0.SelfAttention.k.weight",955            "encoder.block.16.layer.0.SelfAttention.o.weight",956            "encoder.block.16.layer.0.SelfAttention.q.weight",957            "encoder.block.16.layer.0.SelfAttention.v.weight",958            "encoder.block.16.layer.0.layer_norm.weight",959            "encoder.block.16.layer.1.DenseReluDense.wi_0.weight",960            "encoder.block.16.layer.1.DenseReluDense.wi_1.weight",961            "encoder.block.16.layer.1.DenseReluDense.wo.weight",962            "encoder.block.16.layer.1.layer_norm.weight",963            "encoder.block.17.layer.0.SelfAttention.k.weight",964            "encoder.block.17.layer.0.SelfAttention.o.weight",965            "encoder.block.17.layer.0.SelfAttention.q.weight",966            "encoder.block.17.layer.0.SelfAttention.v.weight",967            "encoder.block.17.layer.0.layer_norm.weight",968            "encoder.block.17.layer.1.DenseReluDense.wi_0.weight",969            "encoder.block.17.layer.1.DenseReluDense.wi_1.weight",970            "encoder.block.17.layer.1.DenseReluDense.wo.weight",971            "encoder.block.17.layer.1.layer_norm.weight",972            "encoder.block.18.layer.0.SelfAttention.k.weight",973            "encoder.block.18.layer.0.SelfAttention.o.weight",974            "encoder.block.18.layer.0.SelfAttention.q.weight",975            "encoder.block.18.layer.0.SelfAttention.v.weight",976            "encoder.block.18.layer.0.layer_norm.weight",977            "encoder.block.18.layer.1.DenseReluDense.wi_0.weight",978            "encoder.block.18.layer.1.DenseReluDense.wi_1.weight",979            "encoder.block.18.layer.1.DenseReluDense.wo.weight",980            "encoder.block.18.layer.1.layer_norm.weight",981            "encoder.block.19.layer.0.SelfAttention.k.weight",982            "encoder.block.19.layer.0.SelfAttention.o.weight",983            "encoder.block.19.layer.0.SelfAttention.q.weight",984            "encoder.block.19.layer.0.SelfAttention.v.weight",985            "encoder.block.19.layer.0.layer_norm.weight",986            "encoder.block.19.layer.1.DenseReluDense.wi_0.weight",987            "encoder.block.19.layer.1.DenseReluDense.wi_1.weight",988            "encoder.block.19.layer.1.DenseReluDense.wo.weight",989            "encoder.block.19.layer.1.layer_norm.weight",990            "encoder.block.2.layer.0.SelfAttention.k.weight",991            "encoder.block.2.layer.0.SelfAttention.o.weight",992            "encoder.block.2.layer.0.SelfAttention.q.weight",993            "encoder.block.2.layer.0.SelfAttention.v.weight",994            "encoder.block.2.layer.0.layer_norm.weight",995            "encoder.block.2.layer.1.DenseReluDense.wi_0.weight",996            "encoder.block.2.layer.1.DenseReluDense.wi_1.weight",997            "encoder.block.2.layer.1.DenseReluDense.wo.weight",998            "encoder.block.2.layer.1.layer_norm.weight",999            "encoder.block.20.layer.0.SelfAttention.k.weight",1000            "encoder.block.20.layer.0.SelfAttention.o.weight",1001            "encoder.block.20.layer.0.SelfAttention.q.weight",1002            "encoder.block.20.layer.0.SelfAttention.v.weight",1003            "encoder.block.20.layer.0.layer_norm.weight",1004            "encoder.block.20.layer.1.DenseReluDense.wi_0.weight",1005            "encoder.block.20.layer.1.DenseReluDense.wi_1.weight",1006            "encoder.block.20.layer.1.DenseReluDense.wo.weight",1007            "encoder.block.20.layer.1.layer_norm.weight",1008            "encoder.block.21.layer.0.SelfAttention.k.weight",1009            "encoder.block.21.layer.0.SelfAttention.o.weight",1010            "encoder.block.21.layer.0.SelfAttention.q.weight",1011            "encoder.block.21.layer.0.SelfAttention.v.weight",1012            "encoder.block.21.layer.0.layer_norm.weight",1013            "encoder.block.21.layer.1.DenseReluDense.wi_0.weight",1014            "encoder.block.21.layer.1.DenseReluDense.wi_1.weight",1015            "encoder.block.21.layer.1.DenseReluDense.wo.weight",1016            "encoder.block.21.layer.1.layer_norm.weight",1017            "encoder.block.22.layer.0.SelfAttention.k.weight",1018            "encoder.block.22.layer.0.SelfAttention.o.weight",1019            "encoder.block.22.layer.0.SelfAttention.q.weight",1020            "encoder.block.22.layer.0.SelfAttention.v.weight",1021            "encoder.block.22.layer.0.layer_norm.weight",1022            "encoder.block.22.layer.1.DenseReluDense.wi_0.weight",1023            "encoder.block.22.layer.1.DenseReluDense.wi_1.weight",1024            "encoder.block.22.layer.1.DenseReluDense.wo.weight",1025            "encoder.block.22.layer.1.layer_norm.weight",1026            "encoder.block.23.layer.0.SelfAttention.k.weight",1027            "encoder.block.23.layer.0.SelfAttention.o.weight",1028            "encoder.block.23.layer.0.SelfAttention.q.weight",1029            "encoder.block.23.layer.0.SelfAttention.v.weight",1030            "encoder.block.23.layer.0.layer_norm.weight",1031            "encoder.block.23.layer.1.DenseReluDense.wi_0.weight",1032            "encoder.block.23.layer.1.DenseReluDense.wi_1.weight",1033            "encoder.block.23.layer.1.DenseReluDense.wo.weight",1034            "encoder.block.23.layer.1.layer_norm.weight",1035            "encoder.block.3.layer.0.SelfAttention.k.weight",1036            "encoder.block.3.layer.0.SelfAttention.o.weight",1037            "encoder.block.3.layer.0.SelfAttention.q.weight",1038            "encoder.block.3.layer.0.SelfAttention.v.weight",1039            "encoder.block.3.layer.0.layer_norm.weight",1040            "encoder.block.3.layer.1.DenseReluDense.wi_0.weight",1041            "encoder.block.3.layer.1.DenseReluDense.wi_1.weight",1042            "encoder.block.3.layer.1.DenseReluDense.wo.weight",1043            "encoder.block.3.layer.1.layer_norm.weight",1044            "encoder.block.4.layer.0.SelfAttention.k.weight",1045            "encoder.block.4.layer.0.SelfAttention.o.weight",1046            "encoder.block.4.layer.0.SelfAttention.q.weight",1047            "encoder.block.4.layer.0.SelfAttention.v.weight",1048            "encoder.block.4.layer.0.layer_norm.weight",1049            "encoder.block.4.layer.1.DenseReluDense.wi_0.weight",1050            "encoder.block.4.layer.1.DenseReluDense.wi_1.weight",1051            "encoder.block.4.layer.1.DenseReluDense.wo.weight",1052            "encoder.block.4.layer.1.layer_norm.weight",1053            "encoder.block.5.layer.0.SelfAttention.k.weight",1054            "encoder.block.5.layer.0.SelfAttention.o.weight",1055            "encoder.block.5.layer.0.SelfAttention.q.weight",1056            "encoder.block.5.layer.0.SelfAttention.v.weight",1057            "encoder.block.5.layer.0.layer_norm.weight",1058            "encoder.block.5.layer.1.DenseReluDense.wi_0.weight",1059            "encoder.block.5.layer.1.DenseReluDense.wi_1.weight",1060            "encoder.block.5.layer.1.DenseReluDense.wo.weight",1061            "encoder.block.5.layer.1.layer_norm.weight",1062            "encoder.block.6.layer.0.SelfAttention.k.weight",1063            "encoder.block.6.layer.0.SelfAttention.o.weight",1064            "encoder.block.6.layer.0.SelfAttention.q.weight",1065            "encoder.block.6.layer.0.SelfAttention.v.weight",1066            "encoder.block.6.layer.0.layer_norm.weight",1067            "encoder.block.6.layer.1.DenseReluDense.wi_0.weight",1068            "encoder.block.6.layer.1.DenseReluDense.wi_1.weight",1069            "encoder.block.6.layer.1.DenseReluDense.wo.weight",1070            "encoder.block.6.layer.1.layer_norm.weight",1071            "encoder.block.7.layer.0.SelfAttention.k.weight",1072            "encoder.block.7.layer.0.SelfAttention.o.weight",1073            "encoder.block.7.layer.0.SelfAttention.q.weight",1074            "encoder.block.7.layer.0.SelfAttention.v.weight",1075            "encoder.block.7.layer.0.layer_norm.weight",1076            "encoder.block.7.layer.1.DenseReluDense.wi_0.weight",1077            "encoder.block.7.layer.1.DenseReluDense.wi_1.weight",1078            "encoder.block.7.layer.1.DenseReluDense.wo.weight",1079            "encoder.block.7.layer.1.layer_norm.weight",1080            "encoder.block.8.layer.0.SelfAttention.k.weight",1081            "encoder.block.8.layer.0.SelfAttention.o.weight",1082            "encoder.block.8.layer.0.SelfAttention.q.weight",1083            "encoder.block.8.layer.0.SelfAttention.v.weight",1084            "encoder.block.8.layer.0.layer_norm.weight",1085            "encoder.block.8.layer.1.DenseReluDense.wi_0.weight",1086            "encoder.block.8.layer.1.DenseReluDense.wi_1.weight",1087            "encoder.block.8.layer.1.DenseReluDense.wo.weight",1088            "encoder.block.8.layer.1.layer_norm.weight",1089            "encoder.block.9.layer.0.SelfAttention.k.weight",1090            "encoder.block.9.layer.0.SelfAttention.o.weight",1091            "encoder.block.9.layer.0.SelfAttention.q.weight",1092            "encoder.block.9.layer.0.SelfAttention.v.weight",1093            "encoder.block.9.layer.0.layer_norm.weight",1094            "encoder.block.9.layer.1.DenseReluDense.wi_0.weight",1095            "encoder.block.9.layer.1.DenseReluDense.wi_1.weight",1096            "encoder.block.9.layer.1.DenseReluDense.wo.weight",1097            "encoder.block.9.layer.1.layer_norm.weight",1098            "encoder.embed_tokens.weight",1099            "encoder.final_layer_norm.weight",1100            "shared.weight",1101        ]1102        state_dict_ = {}1103        for name, param in state_dict.items():1104            if name in name_list:1105                state_dict_[name] = param1106        return state_dict_1107 1108