Upload folder using huggingface_hub

README.md

@@ -1,3 +1,50 @@
----
-{}
----
+# RSEdit-DiT-Modified
+
+Local diffusers-compatible checkpoint with custom pipeline code.
+
+## Default Inference Settings
+
+- `torch_dtype=torch.bfloat16`
+- `guidance_scale=4.5`
+- `guidance_interval=(0.0, 1.0)` (default: guidance active for the full denoising schedule)
+- `image_guidance_scale=None` (falls back to `guidance_scale`)
+- `num_inference_steps=50`
+- `clean_caption=False`
+
+## Quick Start
+
+```python
+import torch
+from PIL import Image
+from diffusers import DiffusionPipeline
+
+model_dir = "/data/projects/RSEdit/models/BiliSakura/RSEdit-DiT-Modified"
+img_path = "/data/projects/RSEdit/datasets/BiliSakura/RSCC-RSEdit-Test-Split/images/hurricane-florence_00000109_post_disaster_part1.png"
+out_path = "/data/projects/RSEdit/outputs/hurricane-florence_00000109_rsedit_bf16_cfg4p5_seed12345.png"
+
+prompt = "Severe flooding engulfed the area, submerging all six buildings up to their rooftops, causing partial wall collapses and significant structural weakening. Vegetation along the shoreline was stripped away by rushing waters, exposing bare earth and debris. Roads near the settlement became impassable due to mudslides and erosion, isolating the community. No intact structures remained visible, with every building classified as majorly damaged (Level 2) under disaster protocols."
+
+pipe = DiffusionPipeline.from_pretrained(
+    model_dir,
+    custom_pipeline=f"{model_dir}/pipeline.py",
+    torch_dtype=torch.bfloat16,
+).to("cuda")
+
+print("Pipeline:", pipe.__class__.__name__)
+print("Transformer:", pipe.transformer.__class__.__name__)
+
+image = Image.open(img_path).convert("RGB")
+result = pipe(
+    prompt=prompt,
+    source_image=image,
+    num_inference_steps=50,
+    guidance_scale=4.5,
+    image_guidance_scale=1.5,
+    guidance_interval=(0.0, 1.0),  # default full-range guidance
+    clean_caption=False,
+    generator=torch.Generator(device="cuda").manual_seed(12345),
+).images[0]
+
+result.save(out_path)
+print("Saved:", out_path)
+```

checkpoint-30000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fc99fa72f29f480affe3bddefa65897f47e3ac74f866fe261a735776526fe31
+size 9793028887

checkpoint-30000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3b2b6af81e217acd088d5923355a41fdba319dea5b1a6d62e94d100961a73bc
+size 15537

checkpoint-30000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71a93e7f6752b4aef5269f5c725646a4584b73a104d84e96dba8f0c462ec3477
+size 1465

checkpoint-30000/transformer/config.json

@@ -0,0 +1,33 @@
+{
+  "_class_name": "ConfiguredRSEditModifiedPixArtTransformer2DModel",
+  "_diffusers_version": "0.36.0",
+  "_name_or_path": "/data/models/hf_models/PixArt-alpha/PixArt-XL-2-512x512",
+  "activation_fn": "gelu-approximate",
+  "attention_bias": true,
+  "attention_head_dim": 72,
+  "attention_type": "default",
+  "caption_channels": 4096,
+  "cross_attention_dim": 1152,
+  "double_self_attention": false,
+  "dropout": 0.0,
+  "in_channels": 4,
+  "interpolation_scale": null,
+  "norm_elementwise_affine": false,
+  "norm_eps": 1e-06,
+  "norm_num_groups": 32,
+  "norm_type": "ada_norm_single",
+  "num_attention_heads": 16,
+  "num_embeds_ada_norm": 1000,
+  "num_layers": 28,
+  "num_vector_embeds": null,
+  "only_cross_attention": false,
+  "out_channels": 8,
+  "patch_size": 2,
+  "rsedit_modified_dit": true,
+  "rsedit_window_size": 8,
+  "rsedit_window_skip_every": 1,
+  "sample_size": 64,
+  "upcast_attention": false,
+  "use_additional_conditions": null,
+  "use_linear_projection": false
+}

checkpoint-30000/transformer/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9d3c1f951f58ab1024b6bb4dd9da4331404194573e13949f3d204c4a54e3085
+size 2448140648

logs/rsedit-dit/1777865962.360141/events.out.tfevents.1777865962.ubuntu22.4001.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c5d44f0052443ba2d4d44f13b19749ba557c8fbcead9f08fa9655f26a05597c
+size 3168

logs/rsedit-dit/1777865962.361632/hparams.yml

@@ -0,0 +1,62 @@
+adam_beta1: 0.9
+adam_beta2: 0.99
+adam_epsilon: 1.0e-08
+adam_weight_decay: 0.01
+allow_tf32: false
+cache_dir: null
+center_crop: false
+checkpointing_steps: 10000
+checkpoints_total_limit: 1
+concat_strategy: token
+conditioning_dropout_prob: 0.05
+dataloader_num_workers: 12
+dataset_config_name: null
+dataset_name: null
+edit_prompt_column: edit_prompt
+edited_image_column: edited_image
+enable_xformers_memory_efficient_attention: false
+gradient_accumulation_steps: 4
+gradient_checkpointing: false
+hub_model_id: null
+hub_token: null
+learning_rate: 1.0
+local_rank: 0
+logging_dir: logs
+lr_scheduler: constant
+lr_warmup_steps: 500
+max_grad_norm: 1.0
+max_train_samples: null
+max_train_steps: 30000
+mixed_precision: bf16
+non_ema_revision: null
+num_train_epochs: 16
+num_validation_images: 4
+original_image_column: input_image
+output_dir: /data/projects/RSEdit/models/BiliSakura/RSEdit-DiT-Modified
+pretrained_model_name_or_path: /data/models/hf_models/PixArt-alpha/PixArt-XL-2-512x512
+prodigy_d0: 1.0e-05
+prodigy_d_coef: 1.0
+prodigy_safeguard_warmup: true
+prodigy_use_bias_correction: true
+push_to_hub: false
+random_flip: false
+report_to: tensorboard
+resolution: 512
+resume_from_checkpoint: null
+revision: null
+scale_lr: false
+seed: 42
+train_batch_size: 2
+train_data_dir: /data/data/hf_datasets/BiliSakura/RSCC
+use_8bit_adam: false
+use_ema: true
+use_prodigy: true
+val_annotation_path: /data/data/hf_datasets/BiliSakura/RSCC/RSCC_qvq.jsonl
+val_image_url: null
+val_set_path: /data/data/hf_datasets/BiliSakura/RSCC/val_set.txt
+validation_epochs: 1
+validation_prompt: null
+validation_steps: 0
+variant: null
+window_size: 8
+window_skip_every: 1

logs/rsedit-dit/1777944234.2228265/events.out.tfevents.1777944234.ubuntu22.19227.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9e5eb414ca73533ee3e883a079faa390d9cad9a1cdec862ee9b7f635c7c15da
+size 3232

logs/rsedit-dit/1777944234.2248008/hparams.yml

@@ -0,0 +1,62 @@
+adam_beta1: 0.9
+adam_beta2: 0.99
+adam_epsilon: 1.0e-08
+adam_weight_decay: 0.01
+allow_tf32: false
+cache_dir: null
+center_crop: false
+checkpointing_steps: 10000
+checkpoints_total_limit: 1
+concat_strategy: token
+conditioning_dropout_prob: 0.05
+dataloader_num_workers: 12
+dataset_config_name: null
+dataset_name: null
+edit_prompt_column: edit_prompt
+edited_image_column: edited_image
+enable_xformers_memory_efficient_attention: false
+gradient_accumulation_steps: 4
+gradient_checkpointing: false
+hub_model_id: null
+hub_token: null
+learning_rate: 1.0
+local_rank: 0
+logging_dir: logs
+lr_scheduler: constant
+lr_warmup_steps: 500
+max_grad_norm: 1.0
+max_train_samples: null
+max_train_steps: 30000
+mixed_precision: bf16
+non_ema_revision: null
+num_train_epochs: 16
+num_validation_images: 4
+original_image_column: input_image
+output_dir: /data/projects/RSEdit/models/BiliSakura/RSEdit-DiT-Modified
+pretrained_model_name_or_path: /data/models/hf_models/PixArt-alpha/PixArt-XL-2-512x512
+prodigy_d0: 1.0e-05
+prodigy_d_coef: 1.0
+prodigy_safeguard_warmup: true
+prodigy_use_bias_correction: true
+push_to_hub: false
+random_flip: false
+report_to: tensorboard
+resolution: 512
+resume_from_checkpoint: latest
+revision: null
+scale_lr: false
+seed: 42
+train_batch_size: 2
+train_data_dir: /data/data/hf_datasets/BiliSakura/RSCC
+use_8bit_adam: false
+use_ema: true
+use_prodigy: true
+val_annotation_path: /data/data/hf_datasets/BiliSakura/RSCC/RSCC_qvq.jsonl
+val_image_url: null
+val_set_path: /data/data/hf_datasets/BiliSakura/RSCC/val_set.txt
+validation_epochs: 1
+validation_prompt: null
+validation_steps: 0
+variant: null
+window_size: 8
+window_skip_every: 1

logs/rsedit-dit/events.out.tfevents.1777865962.ubuntu22.4001.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c82e91d9ce95ca392d916d197b88244b6cfbc7045990069b24d470b3a928dbd1
+size 1483578

logs/rsedit-dit/events.out.tfevents.1777944234.ubuntu22.19227.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e41d6f8799d12c8abb581513a93313a20cb30ab4a923411445ac1fd44f1c9a07
+size 88

model_index.json

@@ -0,0 +1,25 @@
+{
+  "_class_name": "RSEditModifiedDiTPipeline",
+  "_diffusers_version": "0.36.0",
+  "_name_or_path": "/data/projects/RSEdit/models/BiliSakura/RSEdit-DiT-Modified",
+  "scheduler": [
+    "diffusers",
+    "DPMSolverMultistepScheduler"
+  ],
+  "text_encoder": [
+    "transformers",
+    "T5EncoderModel"
+  ],
+  "tokenizer": [
+    "transformers",
+    "T5Tokenizer"
+  ],
+  "transformer": [
+    "diffusers",
+    "ConfiguredRSEditModifiedPixArtTransformer2DModel"
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKL"
+  ]
+}

pipeline.py

@@ -0,0 +1,703 @@
+#!/usr/bin/env python
+# coding=utf-8
+
+import os
+from typing import Any, Callable, Dict, List, Optional, Tuple, Union
+
+import PIL.Image
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from safetensors.torch import load_file as safetensors_load_file
+from transformers import T5EncoderModel, T5Tokenizer
+
+from diffusers import AutoencoderKL, PixArtAlphaPipeline
+from diffusers.models.attention import BasicTransformerBlock
+from diffusers.models.modeling_outputs import Transformer2DModelOutput
+from diffusers.models.transformers.pixart_transformer_2d import PixArtTransformer2DModel
+from diffusers.pipelines.pipeline_utils import ImagePipelineOutput
+from diffusers.schedulers import KarrasDiffusionSchedulers
+
+
+EXAMPLE_DOC_STRING = """
+    Examples:
+        ```py
+        >>> import torch
+        >>> from PIL import Image
+        >>> from diffusers import DiffusionPipeline
+
+        >>> pipe = DiffusionPipeline.from_pretrained(
+        ...     "/data/projects/RSEdit/models/BiliSakura/RSEdit-DiT-Modified",
+        ...     custom_pipeline="/data/projects/RSEdit/models/BiliSakura/RSEdit-DiT-Modified/pipeline.py",
+        ...     trust_remote_code=True,
+        ...     torch_dtype=torch.float16
+        ... ).to("cuda")
+
+        >>> source_image = Image.open("satellite_image.png").convert("RGB")
+        >>> image = pipe(
+        ...     prompt="Flood the coastal area",
+        ...     source_image=source_image,
+        ...     num_inference_steps=50,
+        ...     guidance_scale=4.5,
+        ...     guidance_interval=(0.0, 1.0),
+        ... ).images[0]
+        >>> image.save("edited.png")
+        ```
+"""
+
+
+def _window_partition(x: torch.Tensor, window_size: int) -> Tuple[torch.Tensor, Tuple[int, int]]:
+    b, h, w, c = x.shape
+    pad_h = (window_size - h % window_size) % window_size
+    pad_w = (window_size - w % window_size) % window_size
+    if pad_h > 0 or pad_w > 0:
+        x = F.pad(x, (0, 0, 0, pad_w, 0, pad_h))
+    hp, wp = h + pad_h, w + pad_w
+    x = x.view(b, hp // window_size, window_size, wp // window_size, window_size, c)
+    x = x.permute(0, 1, 3, 2, 4, 5).contiguous()
+    x = x.view(-1, window_size * window_size, c)
+    return x, (hp, wp)
+
+
+def _window_unpartition(
+    windows: torch.Tensor,
+    window_size: int,
+    padded_hw: Tuple[int, int],
+    hw: Tuple[int, int],
+) -> torch.Tensor:
+    hp, wp = padded_hw
+    h, w = hw
+    num_windows_per_image = (hp // window_size) * (wp // window_size)
+    b = windows.shape[0] // num_windows_per_image
+    c = windows.shape[-1]
+    x = windows.view(b, hp // window_size, wp // window_size, window_size, window_size, c)
+    x = x.permute(0, 1, 3, 2, 4, 5).contiguous()
+    x = x.view(b, hp, wp, c)
+    if hp > h or wp > w:
+        x = x[:, :h, :w, :].contiguous()
+    return x
+
+
+class DepthwiseConvFeedForward(nn.Module):
+    def __init__(self, original_ff: nn.Module):
+        super().__init__()
+        self.act = original_ff.net[0]
+        self.dropout = original_ff.net[1]
+        self.proj_out = original_ff.net[2]
+        self.final_dropout = original_ff.net[3] if len(original_ff.net) > 3 else None
+        inner_dim = self.proj_out.in_features
+        self.dwconv = nn.Conv2d(inner_dim, inner_dim, kernel_size=3, stride=1, padding=1, groups=inner_dim, bias=False)
+
+    def forward(self, hidden_states: torch.Tensor, height: Optional[int] = None, width: Optional[int] = None) -> torch.Tensor:
+        hidden_states = self.act(hidden_states)
+        if height is not None and width is not None and height * width == hidden_states.shape[1]:
+            bsz, _, channels = hidden_states.shape
+            hidden_states = hidden_states.transpose(1, 2).reshape(bsz, channels, height, width).contiguous()
+            hidden_states = self.dwconv(hidden_states)
+            hidden_states = hidden_states.reshape(bsz, channels, height * width).transpose(1, 2).contiguous()
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.proj_out(hidden_states)
+        if self.final_dropout is not None:
+            hidden_states = self.final_dropout(hidden_states)
+        return hidden_states
+
+
+class ModifiedPixArtTransformerBlock(nn.Module):
+    def __init__(self, block: BasicTransformerBlock, window_size: int = 8):
+        super().__init__()
+        self.norm_type = block.norm_type
+        self.only_cross_attention = block.only_cross_attention
+        self.use_ada_layer_norm_single = block.use_ada_layer_norm_single
+        self.norm1 = block.norm1
+        self.attn1 = block.attn1
+        self.attn2 = block.attn2
+        self.norm2 = block.norm2
+        self.norm3 = getattr(block, "norm3", None)
+        self.ff = DepthwiseConvFeedForward(block.ff)
+        self.fuser = getattr(block, "fuser", None)
+        self.scale_shift_table = getattr(block, "scale_shift_table", None)
+        self.pos_embed = block.pos_embed
+        self.window_size = int(window_size)
+        self._height = None
+        self._width = None
+
+    def set_spatial_shape(self, height: int, width: int):
+        self._height = int(height)
+        self._width = int(width)
+
+    def _apply_window_attention(
+        self,
+        norm_hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor],
+        encoder_hidden_states: Optional[torch.Tensor],
+        cross_attention_kwargs: Dict[str, Any],
+    ) -> torch.Tensor:
+        if (
+            self.window_size <= 0
+            or self._height is None
+            or self._width is None
+            or attention_mask is not None
+            or self.only_cross_attention
+        ):
+            return self.attn1(
+                norm_hidden_states,
+                encoder_hidden_states=encoder_hidden_states if self.only_cross_attention else None,
+                attention_mask=attention_mask,
+                **cross_attention_kwargs,
+            )
+
+        bsz, num_tokens, channels = norm_hidden_states.shape
+        if self._height * self._width != num_tokens:
+            return self.attn1(
+                norm_hidden_states,
+                encoder_hidden_states=encoder_hidden_states if self.only_cross_attention else None,
+                attention_mask=attention_mask,
+                **cross_attention_kwargs,
+            )
+
+        window_input = norm_hidden_states.view(bsz, self._height, self._width, channels)
+        windows, padded_hw = _window_partition(window_input, self.window_size)
+        windows = self.attn1(windows, encoder_hidden_states=None, attention_mask=None, **cross_attention_kwargs)
+        windows = _window_unpartition(windows, self.window_size, padded_hw, (self._height, self._width))
+        return windows.view(bsz, num_tokens, channels)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        timestep: Optional[torch.LongTensor] = None,
+        cross_attention_kwargs: Dict[str, Any] = None,
+        class_labels: Optional[torch.LongTensor] = None,
+        added_cond_kwargs: Optional[Dict[str, torch.Tensor]] = None,
+    ) -> torch.Tensor:
+        batch_size = hidden_states.shape[0]
+
+        if self.norm_type == "ada_norm_single":
+            shift_msa, scale_msa, gate_msa, shift_mlp, scale_mlp, gate_mlp = (
+                self.scale_shift_table[None] + timestep.reshape(batch_size, 6, -1)
+            ).chunk(6, dim=1)
+            norm_hidden_states = self.norm1(hidden_states)
+            norm_hidden_states = norm_hidden_states * (1 + scale_msa) + shift_msa
+        else:
+            norm_hidden_states = self.norm1(hidden_states)
+
+        if self.pos_embed is not None:
+            norm_hidden_states = self.pos_embed(norm_hidden_states)
+
+        cross_attention_kwargs = cross_attention_kwargs.copy() if cross_attention_kwargs is not None else {}
+        gligen_kwargs = cross_attention_kwargs.pop("gligen", None)
+
+        attn_output = self._apply_window_attention(
+            norm_hidden_states=norm_hidden_states,
+            attention_mask=attention_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            cross_attention_kwargs=cross_attention_kwargs,
+        )
+
+        if self.norm_type == "ada_norm_single":
+            attn_output = gate_msa * attn_output
+
+        hidden_states = attn_output + hidden_states
+        if hidden_states.ndim == 4:
+            hidden_states = hidden_states.squeeze(1)
+
+        if gligen_kwargs is not None and self.fuser is not None:
+            hidden_states = self.fuser(hidden_states, gligen_kwargs["objs"])
+
+        if self.attn2 is not None:
+            norm_hidden_states = hidden_states if self.norm_type == "ada_norm_single" else self.norm2(hidden_states)
+            if self.pos_embed is not None and self.norm_type != "ada_norm_single":
+                norm_hidden_states = self.pos_embed(norm_hidden_states)
+            attn_output = self.attn2(
+                norm_hidden_states,
+                encoder_hidden_states=encoder_hidden_states,
+                attention_mask=encoder_attention_mask,
+                **cross_attention_kwargs,
+            )
+            hidden_states = attn_output + hidden_states
+
+        if self.norm_type == "ada_norm_single":
+            norm_hidden_states = self.norm2(hidden_states)
+            norm_hidden_states = norm_hidden_states * (1 + scale_mlp) + shift_mlp
+        else:
+            norm_hidden_states = self.norm3(hidden_states)
+
+        ff_output = self.ff(norm_hidden_states, height=self._height, width=self._width)
+        if self.norm_type == "ada_norm_single":
+            ff_output = gate_mlp * ff_output
+
+        hidden_states = ff_output + hidden_states
+        if hidden_states.ndim == 4:
+            hidden_states = hidden_states.squeeze(1)
+        return hidden_states
+
+
+class RSEditModifiedPixArtTransformer2DModel(PixArtTransformer2DModel):
+    def __init__(
+        self,
+        num_attention_heads: int = 16,
+        attention_head_dim: int = 72,
+        in_channels: int = 4,
+        out_channels: Optional[int] = 8,
+        num_layers: int = 28,
+        dropout: float = 0.0,
+        norm_num_groups: int = 32,
+        cross_attention_dim: Optional[int] = 1152,
+        attention_bias: bool = True,
+        sample_size: int = 128,
+        patch_size: int = 2,
+        activation_fn: str = "gelu-approximate",
+        num_embeds_ada_norm: Optional[int] = 1000,
+        upcast_attention: bool = False,
+        norm_type: str = "ada_norm_single",
+        norm_elementwise_affine: bool = False,
+        norm_eps: float = 1e-6,
+        interpolation_scale: Optional[int] = None,
+        use_additional_conditions: Optional[bool] = None,
+        caption_channels: Optional[int] = None,
+        attention_type: Optional[str] = "default",
+        rsedit_modified_dit: bool = True,
+        rsedit_window_size: int = 8,
+        rsedit_window_skip_every: int = 4,
+        double_self_attention: bool = False,
+        num_vector_embeds: Optional[int] = None,
+        only_cross_attention: bool = False,
+        use_linear_projection: bool = False,
+        **kwargs,
+    ):
+        super().__init__(
+            num_attention_heads=num_attention_heads,
+            attention_head_dim=attention_head_dim,
+            in_channels=in_channels,
+            out_channels=out_channels,
+            num_layers=num_layers,
+            dropout=dropout,
+            norm_num_groups=norm_num_groups,
+            cross_attention_dim=cross_attention_dim,
+            attention_bias=attention_bias,
+            sample_size=sample_size,
+            patch_size=patch_size,
+            activation_fn=activation_fn,
+            num_embeds_ada_norm=num_embeds_ada_norm,
+            upcast_attention=upcast_attention,
+            norm_type=norm_type,
+            norm_elementwise_affine=norm_elementwise_affine,
+            norm_eps=norm_eps,
+            interpolation_scale=interpolation_scale,
+            use_additional_conditions=use_additional_conditions,
+            caption_channels=caption_channels,
+            attention_type=attention_type,
+            **kwargs,
+        )
+        self.register_to_config(
+            rsedit_modified_dit=bool(rsedit_modified_dit),
+            rsedit_window_size=int(rsedit_window_size),
+            rsedit_window_skip_every=int(rsedit_window_skip_every),
+            double_self_attention=bool(double_self_attention),
+            num_vector_embeds=num_vector_embeds,
+            only_cross_attention=bool(only_cross_attention),
+            use_linear_projection=bool(use_linear_projection),
+        )
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: str, *model_args, **kwargs):
+        window_size = kwargs.pop("window_size", None)
+        window_skip_every = kwargs.pop("window_skip_every", None)
+        subfolder = kwargs.get("subfolder")
+        weights_dir = pretrained_model_name_or_path
+        if subfolder:
+            weights_dir = os.path.join(weights_dir, subfolder)
+        safetensors_path = os.path.join(weights_dir, "diffusion_pytorch_model.safetensors")
+
+        # Prefer a direct load path for checkpoints saved from the modified
+        # architecture (keys include ff.act/proj_out/dwconv).
+        if os.path.isfile(safetensors_path):
+            state_dict = safetensors_load_file(safetensors_path)
+            if any(".ff.act." in k or ".ff.dwconv." in k for k in state_dict.keys()):
+                config = cls.load_config(pretrained_model_name_or_path, subfolder=subfolder)
+                model = cls.from_config(config)
+                if window_size is None:
+                    window_size = int(getattr(model.config, "rsedit_window_size", 8))
+                if window_skip_every is None:
+                    window_skip_every = int(getattr(model.config, "rsedit_window_skip_every", 4))
+                model._apply_rsedit_block_modifications(window_size=window_size, window_skip_every=window_skip_every)
+                model.load_state_dict(state_dict, strict=False)
+                torch_dtype = kwargs.get("torch_dtype", None)
+                if torch_dtype is not None:
+                    model = model.to(dtype=torch_dtype)
+                return model
+
+        kwargs.setdefault("low_cpu_mem_usage", False)
+        model = super().from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
+        if window_size is None:
+            window_size = int(getattr(model.config, "rsedit_window_size", 8))
+        if window_skip_every is None:
+            window_skip_every = int(getattr(model.config, "rsedit_window_skip_every", 4))
+        model._apply_rsedit_block_modifications(window_size=window_size, window_skip_every=window_skip_every)
+        return model
+
+    def save_pretrained(self, save_directory: Union[str, os.PathLike], *args, **kwargs):
+        return super().save_pretrained(save_directory, *args, **kwargs)
+
+    def _apply_rsedit_block_modifications(self, window_size: int = 8, window_skip_every: int = 4):
+        if getattr(self, "_rsedit_modified", False):
+            return
+        converted_blocks = []
+        for idx, block in enumerate(self.transformer_blocks):
+            block_window_size = window_size
+            if window_skip_every > 0 and (idx + 1) % window_skip_every == 0:
+                block_window_size = 0
+            converted_blocks.append(ModifiedPixArtTransformerBlock(block, window_size=block_window_size))
+        self.transformer_blocks = nn.ModuleList(converted_blocks)
+        self._rsedit_modified = True
+        self.register_to_config(
+            rsedit_modified_dit=True,
+            rsedit_window_size=int(window_size),
+            rsedit_window_skip_every=int(window_skip_every),
+        )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        timestep: Optional[torch.LongTensor] = None,
+        added_cond_kwargs: Dict[str, torch.Tensor] = None,
+        cross_attention_kwargs: Dict[str, Any] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        return_dict: bool = True,
+    ):
+        if self.use_additional_conditions and added_cond_kwargs is None:
+            raise ValueError("`added_cond_kwargs` cannot be None when using additional conditions for `adaln_single`.")
+        if attention_mask is not None and attention_mask.ndim == 2:
+            attention_mask = (1 - attention_mask.to(hidden_states.dtype)) * -10000.0
+            attention_mask = attention_mask.unsqueeze(1)
+        if encoder_attention_mask is not None and encoder_attention_mask.ndim == 2:
+            encoder_attention_mask = (1 - encoder_attention_mask.to(hidden_states.dtype)) * -10000.0
+            encoder_attention_mask = encoder_attention_mask.unsqueeze(1)
+
+        batch_size = hidden_states.shape[0]
+        height = hidden_states.shape[-2] // self.config.patch_size
+        width = hidden_states.shape[-1] // self.config.patch_size
+        hidden_states = self.pos_embed(hidden_states)
+        timestep, embedded_timestep = self.adaln_single(
+            timestep, added_cond_kwargs, batch_size=batch_size, hidden_dtype=hidden_states.dtype
+        )
+
+        if self.caption_projection is not None:
+            encoder_hidden_states = self.caption_projection(encoder_hidden_states)
+            encoder_hidden_states = encoder_hidden_states.view(batch_size, -1, hidden_states.shape[-1])
+
+        for block in self.transformer_blocks:
+            if hasattr(block, "set_spatial_shape"):
+                block.set_spatial_shape(height, width)
+            if torch.is_grad_enabled() and self.gradient_checkpointing:
+                hidden_states = self._gradient_checkpointing_func(
+                    block,
+                    hidden_states,
+                    attention_mask,
+                    encoder_hidden_states,
+                    encoder_attention_mask,
+                    timestep,
+                    cross_attention_kwargs,
+                    None,
+                )
+            else:
+                hidden_states = block(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    encoder_hidden_states=encoder_hidden_states,
+                    encoder_attention_mask=encoder_attention_mask,
+                    timestep=timestep,
+                    cross_attention_kwargs=cross_attention_kwargs,
+                    class_labels=None,
+                )
+
+        shift, scale = (self.scale_shift_table[None] + embedded_timestep[:, None].to(self.scale_shift_table.device)).chunk(2, dim=1)
+        hidden_states = self.norm_out(hidden_states)
+        hidden_states = hidden_states * (1 + scale.to(hidden_states.device)) + shift.to(hidden_states.device)
+        hidden_states = self.proj_out(hidden_states)
+        hidden_states = hidden_states.squeeze(1)
+
+        hidden_states = hidden_states.reshape((-1, height, width, self.config.patch_size, self.config.patch_size, self.out_channels))
+        hidden_states = torch.einsum("nhwpqc->nchpwq", hidden_states)
+        output = hidden_states.reshape((-1, self.out_channels, height * self.config.patch_size, width * self.config.patch_size))
+        if not return_dict:
+            return (output,)
+        return Transformer2DModelOutput(sample=output)
+
+
+# Keep class name compatible with existing transformer/config.json
+ConfiguredRSEditModifiedPixArtTransformer2DModel = RSEditModifiedPixArtTransformer2DModel
+import diffusers as _diffusers
+setattr(_diffusers, "ConfiguredRSEditModifiedPixArtTransformer2DModel", ConfiguredRSEditModifiedPixArtTransformer2DModel)
+
+
+class RSEditModifiedDiTPipeline(PixArtAlphaPipeline):
+    def __init__(
+        self,
+        vae: AutoencoderKL,
+        text_encoder: T5EncoderModel,
+        tokenizer: T5Tokenizer,
+        transformer: PixArtTransformer2DModel,
+        scheduler: KarrasDiffusionSchedulers,
+    ):
+        super().__init__(
+            vae=vae,
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            transformer=transformer,
+            scheduler=scheduler,
+        )
+
+    def _encode_source_image(
+        self,
+        source_image: PIL.Image.Image,
+        device: torch.device,
+        dtype: torch.dtype,
+        num_images_per_prompt: int = 1,
+    ) -> torch.Tensor:
+        image_np = np.array(source_image.convert("RGB")).astype(np.float32) / 127.5 - 1.0
+        image_tensor = torch.from_numpy(image_np).permute(2, 0, 1).unsqueeze(0)
+        image_tensor = image_tensor.to(device=device, dtype=self.vae.dtype)
+        latents = self.vae.encode(image_tensor).latent_dist.mode()
+        latents = latents * self.vae.config.scaling_factor
+        latents = latents.to(device=device, dtype=dtype)
+        if num_images_per_prompt > 1:
+            latents = latents.repeat(num_images_per_prompt, 1, 1, 1)
+        return latents
+
+    @torch.no_grad()
+    def __call__(
+        self,
+        prompt: Union[str, List[str]] = None,
+        source_image: Union[PIL.Image.Image, List[PIL.Image.Image]] = None,
+        negative_prompt: str = "",
+        num_inference_steps: int = 50,
+        timesteps: List[int] = None,
+        guidance_scale: float = 4.5,
+        guidance_interval: Tuple[float, float] = (0.0, 1.0),
+        image_guidance_scale: Optional[float] = 1.5,
+        num_images_per_prompt: Optional[int] = 1,
+        height: Optional[int] = None,
+        width: Optional[int] = None,
+        eta: float = 0.0,
+        generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
+        latents: Optional[torch.FloatTensor] = None,
+        prompt_embeds: Optional[torch.FloatTensor] = None,
+        prompt_attention_mask: Optional[torch.FloatTensor] = None,
+        negative_prompt_embeds: Optional[torch.FloatTensor] = None,
+        negative_prompt_attention_mask: Optional[torch.FloatTensor] = None,
+        output_type: Optional[str] = "pil",
+        return_dict: bool = True,
+        callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
+        callback_steps: int = 1,
+        clean_caption: bool = True,
+        use_resolution_binning: bool = True,
+        max_sequence_length: int = 120,
+        **kwargs,
+    ) -> Union[ImagePipelineOutput, tuple]:
+        """
+        Run instruction-guided image editing with RSEdit DiT.
+
+        `guidance_interval` controls when classifier-free guidance is active across denoising
+        timesteps. The default `(0.0, 1.0)` enables guidance for the full schedule to match
+        prior behavior.
+        """
+        if source_image is None:
+            raise ValueError("`source_image` must be provided for RSEdit image editing.")
+        if prompt is None and prompt_embeds is None:
+            raise ValueError("Either `prompt` or `prompt_embeds` must be provided.")
+        if len(guidance_interval) != 2 or guidance_interval[0] > guidance_interval[1]:
+            raise ValueError(
+                "`guidance_interval` must be a tuple of two values in ascending order, e.g. (0.0, 1.0)."
+            )
+
+        if height is None:
+            height = self.transformer.config.sample_size * self.vae_scale_factor
+        if width is None:
+            width = self.transformer.config.sample_size * self.vae_scale_factor
+
+        if prompt is not None and isinstance(prompt, str):
+            batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        device = self._execution_device
+
+        if isinstance(source_image, PIL.Image.Image):
+            source_image = source_image.resize((width, height), PIL.Image.LANCZOS)
+        elif isinstance(source_image, list):
+            source_image = [img.resize((width, height), PIL.Image.LANCZOS) for img in source_image]
+            if len(source_image) != batch_size:
+                raise ValueError(f"Number of source images ({len(source_image)}) must match batch size ({batch_size})")
+
+        if isinstance(source_image, list):
+            source_latents_list = []
+            for img in source_image:
+                source_latents_list.append(self._encode_source_image(img, device, self.vae.dtype, num_images_per_prompt).to(device=device))
+            source_latents = torch.cat(source_latents_list, dim=0)
+        else:
+            source_latents = self._encode_source_image(source_image, device, self.vae.dtype, num_images_per_prompt)
+
+        if batch_size > 1 and source_latents.shape[0] == 1:
+            source_latents = source_latents.repeat(batch_size * num_images_per_prompt, 1, 1, 1)
+
+        # Default image_guidance_scale to 1.5 when explicitly unset.
+        if image_guidance_scale is None:
+            image_guidance_scale = 1.5
+        do_classifier_free_guidance = guidance_scale > 1.0 and image_guidance_scale >= 1.0
+
+        (
+            prompt_embeds,
+            prompt_attention_mask,
+            negative_prompt_embeds,
+            negative_prompt_attention_mask,
+        ) = self.encode_prompt(
+            prompt,
+            do_classifier_free_guidance,
+            negative_prompt=negative_prompt,
+            num_images_per_prompt=num_images_per_prompt,
+            device=device,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+            prompt_attention_mask=prompt_attention_mask,
+            negative_prompt_attention_mask=negative_prompt_attention_mask,
+            clean_caption=clean_caption,
+            max_sequence_length=max_sequence_length,
+        )
+
+        base_prompt_embeds = prompt_embeds
+        base_prompt_attention_mask = prompt_attention_mask
+        cfg_prompt_embeds = None
+        cfg_prompt_attention_mask = None
+
+        if do_classifier_free_guidance:
+            cfg_prompt_embeds = torch.cat([prompt_embeds, negative_prompt_embeds, negative_prompt_embeds], dim=0)
+            cfg_prompt_attention_mask = torch.cat(
+                [prompt_attention_mask, negative_prompt_attention_mask, negative_prompt_attention_mask], dim=0
+            )
+
+        self.scheduler.set_timesteps(num_inference_steps, device=device)
+        timesteps = self.scheduler.timesteps
+
+        num_channels_latents = self.transformer.config.in_channels
+        latents = self.prepare_latents(
+            batch_size * num_images_per_prompt,
+            num_channels_latents,
+            height,
+            width,
+            prompt_embeds.dtype,
+            device,
+            generator,
+            latents,
+        )
+
+        source_latents = source_latents.to(device=latents.device)
+        extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
+
+        base_added_cond_kwargs = {"resolution": None, "aspect_ratio": None}
+        cfg_added_cond_kwargs = base_added_cond_kwargs
+        use_additional_conditions = bool(
+            getattr(getattr(self.transformer, "adaln_single", None), "emb", None) is not None
+            and getattr(self.transformer.adaln_single.emb, "use_additional_conditions", False)
+        )
+        if use_additional_conditions:
+            resolution = torch.tensor([height, width]).repeat(batch_size * num_images_per_prompt, 1)
+            aspect_ratio = torch.tensor([float(height / width)]).repeat(batch_size * num_images_per_prompt, 1)
+            resolution = resolution.to(dtype=prompt_embeds.dtype, device=device)
+            aspect_ratio = aspect_ratio.to(dtype=prompt_embeds.dtype, device=device)
+            if do_classifier_free_guidance:
+                cfg_added_cond_kwargs = {
+                    "resolution": torch.cat([resolution, resolution, resolution], dim=0),
+                    "aspect_ratio": torch.cat([aspect_ratio, aspect_ratio, aspect_ratio], dim=0),
+                }
+            base_added_cond_kwargs = {"resolution": resolution, "aspect_ratio": aspect_ratio}
+
+        num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
+        with self.progress_bar(total=num_inference_steps) as progress_bar:
+            for i, t in enumerate(timesteps):
+                guidance_active = do_classifier_free_guidance and guidance_interval[0] <= float(t) <= guidance_interval[1]
+
+                latent_model_input = torch.cat([latents] * 3) if guidance_active else latents
+                latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
+                if guidance_active:
+                    source_latents_input = torch.cat([source_latents, source_latents, torch.zeros_like(source_latents)], dim=0)
+                    prompt_embeds_input = cfg_prompt_embeds
+                    prompt_attention_mask_input = cfg_prompt_attention_mask
+                    added_cond_kwargs_input = cfg_added_cond_kwargs
+                else:
+                    source_latents_input = source_latents
+                    prompt_embeds_input = base_prompt_embeds
+                    prompt_attention_mask_input = base_prompt_attention_mask
+                    added_cond_kwargs_input = base_added_cond_kwargs
+
+                source_latents_input = source_latents_input.to(device=latent_model_input.device)
+                concatenated_latents = torch.cat([source_latents_input, latent_model_input], dim=3)
+
+                current_timestep = t
+                if not torch.is_tensor(current_timestep):
+                    is_mps = concatenated_latents.device.type == "mps"
+                    is_npu = concatenated_latents.device.type == "npu"
+                    if isinstance(current_timestep, float):
+                        dtype = torch.float32 if (is_mps or is_npu) else torch.float64
+                    else:
+                        dtype = torch.int32 if (is_mps or is_npu) else torch.int64
+                    current_timestep = torch.tensor([current_timestep], dtype=dtype, device=concatenated_latents.device)
+                elif len(current_timestep.shape) == 0:
+                    current_timestep = current_timestep[None].to(concatenated_latents.device)
+                current_timestep = current_timestep.expand(concatenated_latents.shape[0])
+
+                noise_pred = self.transformer(
+                    concatenated_latents,
+                    encoder_hidden_states=prompt_embeds_input,
+                    encoder_attention_mask=prompt_attention_mask_input,
+                    timestep=current_timestep,
+                    added_cond_kwargs=added_cond_kwargs_input,
+                    return_dict=False,
+                )[0]
+
+                target_width = latents.shape[3]
+                noise_pred = noise_pred[:, :, :, target_width:]
+                if noise_pred.shape[1] == 2 * num_channels_latents:
+                    noise_pred, _ = noise_pred.chunk(2, dim=1)
+
+                if guidance_active:
+                    # noise_pred batch: [Text+Image, Image, None]
+                    noise_pred_text, noise_pred_image, noise_pred_uncond = noise_pred.chunk(3)
+
+                    # IP2P CFG Formula:
+                    # pred = uncond + s_text * (text - image) + s_image * (image - uncond)
+                    # Mapping:
+                    # e(c_I, c_T) -> noise_pred_text (Full)
+                    # e(c_I, phi) -> noise_pred_image (Image only, Null Text)
+                    # e(phi, phi) -> noise_pred_uncond (Unconditional)
+                    noise_pred = (
+                        noise_pred_uncond
+                        + guidance_scale * (noise_pred_text - noise_pred_image)
+                        + image_guidance_scale * (noise_pred_image - noise_pred_uncond)
+                    )
+
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
+
+                if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
+                    progress_bar.update()
+                    if callback is not None and i % callback_steps == 0:
+                        step_idx = i // getattr(self.scheduler, "order", 1)
+                        callback(step_idx, t, latents)
+
+        if output_type != "latent":
+            image = self.vae.decode(latents.to(self.vae.dtype) / self.vae.config.scaling_factor, return_dict=False)[0]
+            image = self.image_processor.postprocess(image, output_type=output_type)
+        else:
+            image = latents
+
+        self.maybe_free_model_hooks()
+        if not return_dict:
+            return (image,)
+        return ImagePipelineOutput(images=image)

scheduler/scheduler_config.json

@@ -0,0 +1,32 @@
+{
+  "_class_name": "DPMSolverMultistepScheduler",
+  "_diffusers_version": "0.36.0.dev0",
+  "algorithm_type": "dpmsolver++",
+  "beta_end": 0.02,
+  "beta_schedule": "linear",
+  "beta_start": 0.0001,
+  "dynamic_thresholding_ratio": 0.995,
+  "euler_at_final": false,
+  "final_sigmas_type": "zero",
+  "flow_shift": 1.0,
+  "lambda_min_clipped": -Infinity,
+  "lower_order_final": true,
+  "num_train_timesteps": 1000,
+  "prediction_type": "epsilon",
+  "rescale_betas_zero_snr": false,
+  "sample_max_value": 1.0,
+  "solver_order": 2,
+  "solver_type": "midpoint",
+  "steps_offset": 0,
+  "thresholding": false,
+  "time_shift_type": "exponential",
+  "timestep_spacing": "linspace",
+  "trained_betas": null,
+  "use_beta_sigmas": false,
+  "use_dynamic_shifting": false,
+  "use_exponential_sigmas": false,
+  "use_flow_sigmas": false,
+  "use_karras_sigmas": false,
+  "use_lu_lambdas": false,
+  "variance_type": null
+}

text_encoder/config.json

@@ -0,0 +1,32 @@
+{
+  "architectures": [
+    "T5EncoderModel"
+  ],
+  "classifier_dropout": 0.0,
+  "d_ff": 10240,
+  "d_kv": 64,
+  "d_model": 4096,
+  "decoder_start_token_id": 0,
+  "dense_act_fn": "gelu_new",
+  "dropout_rate": 0.1,
+  "dtype": "bfloat16",
+  "eos_token_id": 1,
+  "feed_forward_proj": "gated-gelu",
+  "initializer_factor": 1.0,
+  "is_encoder_decoder": false,
+  "is_gated_act": true,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "t5",
+  "num_decoder_layers": 24,
+  "num_heads": 64,
+  "num_layers": 24,
+  "output_past": true,
+  "pad_token_id": 0,
+  "relative_attention_max_distance": 128,
+  "relative_attention_num_buckets": 32,
+  "tie_encoder_decoder": true,
+  "tie_word_embeddings": false,
+  "transformers_version": "5.0.0.dev0",
+  "use_cache": false,
+  "vocab_size": 32128
+}

text_encoder/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec87bffd1923e8b2774a6d240c922a41f6143081d52cf83b8fe39e9d838c893e
+size 4994582224

text_encoder/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5640855b301fcdbceddfa90ae8066cd9414aff020552a201a255ecf2059da00
+size 4530066360

text_encoder/model.safetensors.index.json

@@ -0,0 +1,227 @@
+{
+  "metadata": {
+    "total_parameters": 4762310656,
+    "total_size": 9524621312
+  },
+  "weight_map": {
+    "encoder.block.0.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.0.SelfAttention.relative_attention_bias.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.0.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.1.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.10.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.11.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.12.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.12.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.12.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.12.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.12.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.12.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.12.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.12.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.12.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.13.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.14.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.15.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.16.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.17.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.18.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.19.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.2.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.2.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.20.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.20.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.21.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.22.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.0.SelfAttention.k.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.0.SelfAttention.o.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.0.SelfAttention.q.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.0.SelfAttention.v.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.0.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.1.DenseReluDense.wi_0.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.1.DenseReluDense.wi_1.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.1.DenseReluDense.wo.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.23.layer.1.layer_norm.weight": "model-00002-of-00002.safetensors",
+    "encoder.block.3.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.3.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.4.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.5.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.6.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.7.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.8.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.0.SelfAttention.k.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.0.SelfAttention.o.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.0.SelfAttention.q.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.0.SelfAttention.v.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.0.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.1.DenseReluDense.wi_0.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.1.DenseReluDense.wi_1.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.1.DenseReluDense.wo.weight": "model-00001-of-00002.safetensors",
+    "encoder.block.9.layer.1.layer_norm.weight": "model-00001-of-00002.safetensors",
+    "encoder.final_layer_norm.weight": "model-00002-of-00002.safetensors",
+    "shared.weight": "model-00001-of-00002.safetensors"
+  }
+}

tokenizer/added_tokens.json

@@ -0,0 +1,102 @@
+{
+  "<extra_id_0>": 32099,
+  "<extra_id_10>": 32089,
+  "<extra_id_11>": 32088,
+  "<extra_id_12>": 32087,
+  "<extra_id_13>": 32086,
+  "<extra_id_14>": 32085,
+  "<extra_id_15>": 32084,
+  "<extra_id_16>": 32083,
+  "<extra_id_17>": 32082,
+  "<extra_id_18>": 32081,
+  "<extra_id_19>": 32080,
+  "<extra_id_1>": 32098,
+  "<extra_id_20>": 32079,
+  "<extra_id_21>": 32078,
+  "<extra_id_22>": 32077,
+  "<extra_id_23>": 32076,
+  "<extra_id_24>": 32075,
+  "<extra_id_25>": 32074,
+  "<extra_id_26>": 32073,
+  "<extra_id_27>": 32072,
+  "<extra_id_28>": 32071,
+  "<extra_id_29>": 32070,
+  "<extra_id_2>": 32097,
+  "<extra_id_30>": 32069,
+  "<extra_id_31>": 32068,
+  "<extra_id_32>": 32067,
+  "<extra_id_33>": 32066,
+  "<extra_id_34>": 32065,
+  "<extra_id_35>": 32064,
+  "<extra_id_36>": 32063,
+  "<extra_id_37>": 32062,
+  "<extra_id_38>": 32061,
+  "<extra_id_39>": 32060,
+  "<extra_id_3>": 32096,
+  "<extra_id_40>": 32059,
+  "<extra_id_41>": 32058,
+  "<extra_id_42>": 32057,
+  "<extra_id_43>": 32056,
+  "<extra_id_44>": 32055,
+  "<extra_id_45>": 32054,
+  "<extra_id_46>": 32053,
+  "<extra_id_47>": 32052,
+  "<extra_id_48>": 32051,
+  "<extra_id_49>": 32050,
+  "<extra_id_4>": 32095,
+  "<extra_id_50>": 32049,
+  "<extra_id_51>": 32048,
+  "<extra_id_52>": 32047,
+  "<extra_id_53>": 32046,
+  "<extra_id_54>": 32045,
+  "<extra_id_55>": 32044,
+  "<extra_id_56>": 32043,
+  "<extra_id_57>": 32042,
+  "<extra_id_58>": 32041,
+  "<extra_id_59>": 32040,
+  "<extra_id_5>": 32094,
+  "<extra_id_60>": 32039,
+  "<extra_id_61>": 32038,
+  "<extra_id_62>": 32037,
+  "<extra_id_63>": 32036,
+  "<extra_id_64>": 32035,
+  "<extra_id_65>": 32034,
+  "<extra_id_66>": 32033,
+  "<extra_id_67>": 32032,
+  "<extra_id_68>": 32031,
+  "<extra_id_69>": 32030,
+  "<extra_id_6>": 32093,
+  "<extra_id_70>": 32029,
+  "<extra_id_71>": 32028,
+  "<extra_id_72>": 32027,
+  "<extra_id_73>": 32026,
+  "<extra_id_74>": 32025,
+  "<extra_id_75>": 32024,
+  "<extra_id_76>": 32023,
+  "<extra_id_77>": 32022,
+  "<extra_id_78>": 32021,
+  "<extra_id_79>": 32020,
+  "<extra_id_7>": 32092,
+  "<extra_id_80>": 32019,
+  "<extra_id_81>": 32018,
+  "<extra_id_82>": 32017,
+  "<extra_id_83>": 32016,
+  "<extra_id_84>": 32015,
+  "<extra_id_85>": 32014,
+  "<extra_id_86>": 32013,
+  "<extra_id_87>": 32012,
+  "<extra_id_88>": 32011,
+  "<extra_id_89>": 32010,
+  "<extra_id_8>": 32091,
+  "<extra_id_90>": 32009,
+  "<extra_id_91>": 32008,
+  "<extra_id_92>": 32007,
+  "<extra_id_93>": 32006,
+  "<extra_id_94>": 32005,
+  "<extra_id_95>": 32004,
+  "<extra_id_96>": 32003,
+  "<extra_id_97>": 32002,
+  "<extra_id_98>": 32001,
+  "<extra_id_99>": 32000,
+  "<extra_id_9>": 32090
+}

tokenizer/special_tokens_map.json

@@ -0,0 +1,125 @@
+{
+  "additional_special_tokens": [
+    "<extra_id_0>",
+    "<extra_id_1>",
+    "<extra_id_2>",
+    "<extra_id_3>",
+    "<extra_id_4>",
+    "<extra_id_5>",
+    "<extra_id_6>",
+    "<extra_id_7>",
+    "<extra_id_8>",
+    "<extra_id_9>",
+    "<extra_id_10>",
+    "<extra_id_11>",
+    "<extra_id_12>",
+    "<extra_id_13>",
+    "<extra_id_14>",
+    "<extra_id_15>",
+    "<extra_id_16>",
+    "<extra_id_17>",
+    "<extra_id_18>",
+    "<extra_id_19>",
+    "<extra_id_20>",
+    "<extra_id_21>",
+    "<extra_id_22>",
+    "<extra_id_23>",
+    "<extra_id_24>",
+    "<extra_id_25>",
+    "<extra_id_26>",
+    "<extra_id_27>",
+    "<extra_id_28>",
+    "<extra_id_29>",
+    "<extra_id_30>",
+    "<extra_id_31>",
+    "<extra_id_32>",
+    "<extra_id_33>",
+    "<extra_id_34>",
+    "<extra_id_35>",
+    "<extra_id_36>",
+    "<extra_id_37>",
+    "<extra_id_38>",
+    "<extra_id_39>",
+    "<extra_id_40>",
+    "<extra_id_41>",
+    "<extra_id_42>",
+    "<extra_id_43>",
+    "<extra_id_44>",
+    "<extra_id_45>",
+    "<extra_id_46>",
+    "<extra_id_47>",
+    "<extra_id_48>",
+    "<extra_id_49>",
+    "<extra_id_50>",
+    "<extra_id_51>",
+    "<extra_id_52>",
+    "<extra_id_53>",
+    "<extra_id_54>",
+    "<extra_id_55>",
+    "<extra_id_56>",
+    "<extra_id_57>",
+    "<extra_id_58>",
+    "<extra_id_59>",
+    "<extra_id_60>",
+    "<extra_id_61>",
+    "<extra_id_62>",
+    "<extra_id_63>",
+    "<extra_id_64>",
+    "<extra_id_65>",
+    "<extra_id_66>",
+    "<extra_id_67>",
+    "<extra_id_68>",
+    "<extra_id_69>",
+    "<extra_id_70>",
+    "<extra_id_71>",
+    "<extra_id_72>",
+    "<extra_id_73>",
+    "<extra_id_74>",
+    "<extra_id_75>",
+    "<extra_id_76>",
+    "<extra_id_77>",
+    "<extra_id_78>",
+    "<extra_id_79>",
+    "<extra_id_80>",
+    "<extra_id_81>",
+    "<extra_id_82>",
+    "<extra_id_83>",
+    "<extra_id_84>",
+    "<extra_id_85>",
+    "<extra_id_86>",
+    "<extra_id_87>",
+    "<extra_id_88>",
+    "<extra_id_89>",
+    "<extra_id_90>",
+    "<extra_id_91>",
+    "<extra_id_92>",
+    "<extra_id_93>",
+    "<extra_id_94>",
+    "<extra_id_95>",
+    "<extra_id_96>",
+    "<extra_id_97>",
+    "<extra_id_98>",
+    "<extra_id_99>"
+  ],
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer/spiece.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d60acb128cf7b7f2536e8f38a5b18a05535c9e14c7a355904270e15b0945ea86
+size 791656

tokenizer/tokenizer_config.json

@@ -0,0 +1,941 @@
+{
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32000": {
+      "content": "<extra_id_99>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32001": {
+      "content": "<extra_id_98>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32002": {
+      "content": "<extra_id_97>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32003": {
+      "content": "<extra_id_96>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32004": {
+      "content": "<extra_id_95>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32005": {
+      "content": "<extra_id_94>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32006": {
+      "content": "<extra_id_93>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32007": {
+      "content": "<extra_id_92>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32008": {
+      "content": "<extra_id_91>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32009": {
+      "content": "<extra_id_90>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32010": {
+      "content": "<extra_id_89>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32011": {
+      "content": "<extra_id_88>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32012": {
+      "content": "<extra_id_87>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32013": {
+      "content": "<extra_id_86>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32014": {
+      "content": "<extra_id_85>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32015": {
+      "content": "<extra_id_84>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32016": {
+      "content": "<extra_id_83>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32017": {
+      "content": "<extra_id_82>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32018": {
+      "content": "<extra_id_81>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32019": {
+      "content": "<extra_id_80>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32020": {
+      "content": "<extra_id_79>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32021": {
+      "content": "<extra_id_78>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32022": {
+      "content": "<extra_id_77>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32023": {
+      "content": "<extra_id_76>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32024": {
+      "content": "<extra_id_75>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32025": {
+      "content": "<extra_id_74>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32026": {
+      "content": "<extra_id_73>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32027": {
+      "content": "<extra_id_72>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32028": {
+      "content": "<extra_id_71>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32029": {
+      "content": "<extra_id_70>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32030": {
+      "content": "<extra_id_69>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32031": {
+      "content": "<extra_id_68>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32032": {
+      "content": "<extra_id_67>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32033": {
+      "content": "<extra_id_66>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32034": {
+      "content": "<extra_id_65>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32035": {
+      "content": "<extra_id_64>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32036": {
+      "content": "<extra_id_63>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32037": {
+      "content": "<extra_id_62>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32038": {
+      "content": "<extra_id_61>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32039": {
+      "content": "<extra_id_60>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32040": {
+      "content": "<extra_id_59>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32041": {
+      "content": "<extra_id_58>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32042": {
+      "content": "<extra_id_57>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32043": {
+      "content": "<extra_id_56>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32044": {
+      "content": "<extra_id_55>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32045": {
+      "content": "<extra_id_54>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32046": {
+      "content": "<extra_id_53>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32047": {
+      "content": "<extra_id_52>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32048": {
+      "content": "<extra_id_51>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32049": {
+      "content": "<extra_id_50>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32050": {
+      "content": "<extra_id_49>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32051": {
+      "content": "<extra_id_48>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32052": {
+      "content": "<extra_id_47>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32053": {
+      "content": "<extra_id_46>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32054": {
+      "content": "<extra_id_45>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32055": {
+      "content": "<extra_id_44>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32056": {
+      "content": "<extra_id_43>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32057": {
+      "content": "<extra_id_42>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32058": {
+      "content": "<extra_id_41>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32059": {
+      "content": "<extra_id_40>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32060": {
+      "content": "<extra_id_39>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32061": {
+      "content": "<extra_id_38>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32062": {
+      "content": "<extra_id_37>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32063": {
+      "content": "<extra_id_36>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32064": {
+      "content": "<extra_id_35>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32065": {
+      "content": "<extra_id_34>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32066": {
+      "content": "<extra_id_33>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32067": {
+      "content": "<extra_id_32>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32068": {
+      "content": "<extra_id_31>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32069": {
+      "content": "<extra_id_30>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32070": {
+      "content": "<extra_id_29>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32071": {
+      "content": "<extra_id_28>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32072": {
+      "content": "<extra_id_27>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32073": {
+      "content": "<extra_id_26>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32074": {
+      "content": "<extra_id_25>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32075": {
+      "content": "<extra_id_24>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32076": {
+      "content": "<extra_id_23>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32077": {
+      "content": "<extra_id_22>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32078": {
+      "content": "<extra_id_21>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32079": {
+      "content": "<extra_id_20>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32080": {
+      "content": "<extra_id_19>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32081": {
+      "content": "<extra_id_18>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32082": {
+      "content": "<extra_id_17>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32083": {
+      "content": "<extra_id_16>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32084": {
+      "content": "<extra_id_15>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32085": {
+      "content": "<extra_id_14>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32086": {
+      "content": "<extra_id_13>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32087": {
+      "content": "<extra_id_12>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32088": {
+      "content": "<extra_id_11>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32089": {
+      "content": "<extra_id_10>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32090": {
+      "content": "<extra_id_9>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32091": {
+      "content": "<extra_id_8>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32092": {
+      "content": "<extra_id_7>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32093": {
+      "content": "<extra_id_6>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32094": {
+      "content": "<extra_id_5>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32095": {
+      "content": "<extra_id_4>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32096": {
+      "content": "<extra_id_3>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32097": {
+      "content": "<extra_id_2>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32098": {
+      "content": "<extra_id_1>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    },
+    "32099": {
+      "content": "<extra_id_0>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": true,
+      "single_word": true,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<extra_id_0>",
+    "<extra_id_1>",
+    "<extra_id_2>",
+    "<extra_id_3>",
+    "<extra_id_4>",
+    "<extra_id_5>",
+    "<extra_id_6>",
+    "<extra_id_7>",
+    "<extra_id_8>",
+    "<extra_id_9>",
+    "<extra_id_10>",
+    "<extra_id_11>",
+    "<extra_id_12>",
+    "<extra_id_13>",
+    "<extra_id_14>",
+    "<extra_id_15>",
+    "<extra_id_16>",
+    "<extra_id_17>",
+    "<extra_id_18>",
+    "<extra_id_19>",
+    "<extra_id_20>",
+    "<extra_id_21>",
+    "<extra_id_22>",
+    "<extra_id_23>",
+    "<extra_id_24>",
+    "<extra_id_25>",
+    "<extra_id_26>",
+    "<extra_id_27>",
+    "<extra_id_28>",
+    "<extra_id_29>",
+    "<extra_id_30>",
+    "<extra_id_31>",
+    "<extra_id_32>",
+    "<extra_id_33>",
+    "<extra_id_34>",
+    "<extra_id_35>",
+    "<extra_id_36>",
+    "<extra_id_37>",
+    "<extra_id_38>",
+    "<extra_id_39>",
+    "<extra_id_40>",
+    "<extra_id_41>",
+    "<extra_id_42>",
+    "<extra_id_43>",
+    "<extra_id_44>",
+    "<extra_id_45>",
+    "<extra_id_46>",
+    "<extra_id_47>",
+    "<extra_id_48>",
+    "<extra_id_49>",
+    "<extra_id_50>",
+    "<extra_id_51>",
+    "<extra_id_52>",
+    "<extra_id_53>",
+    "<extra_id_54>",
+    "<extra_id_55>",
+    "<extra_id_56>",
+    "<extra_id_57>",
+    "<extra_id_58>",
+    "<extra_id_59>",
+    "<extra_id_60>",
+    "<extra_id_61>",
+    "<extra_id_62>",
+    "<extra_id_63>",
+    "<extra_id_64>",
+    "<extra_id_65>",
+    "<extra_id_66>",
+    "<extra_id_67>",
+    "<extra_id_68>",
+    "<extra_id_69>",
+    "<extra_id_70>",
+    "<extra_id_71>",
+    "<extra_id_72>",
+    "<extra_id_73>",
+    "<extra_id_74>",
+    "<extra_id_75>",
+    "<extra_id_76>",
+    "<extra_id_77>",
+    "<extra_id_78>",
+    "<extra_id_79>",
+    "<extra_id_80>",
+    "<extra_id_81>",
+    "<extra_id_82>",
+    "<extra_id_83>",
+    "<extra_id_84>",
+    "<extra_id_85>",
+    "<extra_id_86>",
+    "<extra_id_87>",
+    "<extra_id_88>",
+    "<extra_id_89>",
+    "<extra_id_90>",
+    "<extra_id_91>",
+    "<extra_id_92>",
+    "<extra_id_93>",
+    "<extra_id_94>",
+    "<extra_id_95>",
+    "<extra_id_96>",
+    "<extra_id_97>",
+    "<extra_id_98>",
+    "<extra_id_99>"
+  ],
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "</s>",
+  "extra_ids": 100,
+  "extra_special_tokens": {},
+  "legacy": true,
+  "model_max_length": 512,
+  "pad_token": "<pad>",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "T5Tokenizer",
+  "unk_token": "<unk>"
+}

transformer/config.json

@@ -0,0 +1,33 @@
+{
+  "_class_name": "RSEditModifiedPixArtTransformer2DModel",
+  "_diffusers_version": "0.36.0",
+  "_name_or_path": "/data/models/hf_models/PixArt-alpha/PixArt-XL-2-512x512",
+  "activation_fn": "gelu-approximate",
+  "attention_bias": true,
+  "attention_head_dim": 72,
+  "attention_type": "default",
+  "caption_channels": 4096,
+  "cross_attention_dim": 1152,
+  "double_self_attention": false,
+  "dropout": 0.0,
+  "in_channels": 4,
+  "interpolation_scale": null,
+  "norm_elementwise_affine": false,
+  "norm_eps": 1e-06,
+  "norm_num_groups": 32,
+  "norm_type": "ada_norm_single",
+  "num_attention_heads": 16,
+  "num_embeds_ada_norm": 1000,
+  "num_layers": 28,
+  "num_vector_embeds": null,
+  "only_cross_attention": false,
+  "out_channels": 8,
+  "patch_size": 2,
+  "rsedit_modified_dit": true,
+  "rsedit_window_size": 8,
+  "rsedit_window_skip_every": 1,
+  "sample_size": 64,
+  "upcast_attention": false,
+  "use_additional_conditions": null,
+  "use_linear_projection": false
+}

transformer/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9d3c1f951f58ab1024b6bb4dd9da4331404194573e13949f3d204c4a54e3085
+size 2448140648

vae/config.json

@@ -0,0 +1,38 @@
+{
+  "_class_name": "AutoencoderKL",
+  "_diffusers_version": "0.36.0.dev0",
+  "_name_or_path": "/data/models/hf_models/PixArt-alpha/PixArt-XL-2-512x512",
+  "act_fn": "silu",
+  "block_out_channels": [
+    128,
+    256,
+    512,
+    512
+  ],
+  "down_block_types": [
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D",
+    "DownEncoderBlock2D"
+  ],
+  "force_upcast": true,
+  "in_channels": 3,
+  "latent_channels": 4,
+  "latents_mean": null,
+  "latents_std": null,
+  "layers_per_block": 2,
+  "mid_block_add_attention": true,
+  "norm_num_groups": 32,
+  "out_channels": 3,
+  "sample_size": 256,
+  "scaling_factor": 0.18215,
+  "shift_factor": null,
+  "up_block_types": [
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D",
+    "UpDecoderBlock2D"
+  ],
+  "use_post_quant_conv": true,
+  "use_quant_conv": true
+}

vae/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d6de79b2229c59391de90d1a27aa62f4cc5423cc82621ad2307206007fa965e
+size 167335590