Update seed2_tokenizer.py

seed2_tokenizer.py

@@ -36,7 +36,8 @@ from diffusers import DiffusionPipeline
 from PIL import Image
 from torchvision import transforms
 
-WEIGHTS_NAME = 'seed_quantizer.pt'
+import torch.utils.checkpoint as checkpoint
+
 DIFFUSION_NAME = 'stabilityai/stable-diffusion-2-1-unclip'
 
 # from qformer.qformer_quantizer import Blip2QformerQuantizer
@@ -66,8 +67,8 @@ import torch.nn as nn
 import torch.distributed as dist
 import torch.nn.functional as F
 
+from timm.models.layers import drop_path, to_2tuple, trunc_normal_
 
-from .eva_vit import create_eva_vit_g, VisionTransformerEvaClip
 from transformers import BertTokenizer
 
 import math
@@ -130,6 +131,420 @@ from transformers.models.bert.configuration_bert import BertConfig
 
 #torch.set_printoptions(profile="full")
 
+class DropPathEvaVit(nn.Module):
+    """Drop paths (Stochastic Depth) per sample  (when applied in main path of residual blocks).
+    """
+    def __init__(self, drop_prob=None):
+        super(DropPathEvaVit, self).__init__()
+        self.drop_prob = drop_prob
+
+    def forward(self, x):
+        return drop_path(x, self.drop_prob, self.training)
+
+    def extra_repr(self) -> str:
+        return 'p={}'.format(self.drop_prob)
+
+
+class MlpEvaVit(nn.Module):
+    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):
+        super().__init__()
+        out_features = out_features or in_features
+        hidden_features = hidden_features or in_features
+        self.fc1 = nn.Linear(in_features, hidden_features)
+        self.act = act_layer()
+        self.fc2 = nn.Linear(hidden_features, out_features)
+        self.drop = nn.Dropout(drop)
+
+    def forward(self, x):
+        x = self.fc1(x)
+        x = self.act(x)
+        # x = self.drop(x)
+        # commit this for the orignal BERT implement
+        x = self.fc2(x)
+        x = self.drop(x)
+        return x
+
+
+class AttentionEvaVit(nn.Module):
+    def __init__(self,
+                 dim,
+                 num_heads=8,
+                 qkv_bias=False,
+                 qk_scale=None,
+                 attn_drop=0.,
+                 proj_drop=0.,
+                 window_size=None,
+                 attn_head_dim=None):
+        super().__init__()
+        self.num_heads = num_heads
+        head_dim = dim // num_heads
+        if attn_head_dim is not None:
+            head_dim = attn_head_dim
+        all_head_dim = head_dim * self.num_heads
+        self.scale = qk_scale or head_dim**-0.5
+
+        self.qkv = nn.Linear(dim, all_head_dim * 3, bias=False)
+        if qkv_bias:
+            self.q_bias = nn.Parameter(torch.zeros(all_head_dim))
+            self.v_bias = nn.Parameter(torch.zeros(all_head_dim))
+        else:
+            self.q_bias = None
+            self.v_bias = None
+
+        if window_size:
+            self.window_size = window_size
+            self.num_relative_distance = (2 * window_size[0] - 1) * (2 * window_size[1] - 1) + 3
+            self.relative_position_bias_table = nn.Parameter(torch.zeros(self.num_relative_distance,
+                                                                         num_heads))  # 2*Wh-1 * 2*Ww-1, nH
+            # cls to token & token 2 cls & cls to cls
+
+            # get pair-wise relative position index for each token inside the window
+            coords_h = torch.arange(window_size[0])
+            coords_w = torch.arange(window_size[1])
+            coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
+            coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
+            relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
+            relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
+            relative_coords[:, :, 0] += window_size[0] - 1  # shift to start from 0
+            relative_coords[:, :, 1] += window_size[1] - 1
+            relative_coords[:, :, 0] *= 2 * window_size[1] - 1
+            relative_position_index = \
+                torch.zeros(size=(window_size[0] * window_size[1] + 1, ) * 2, dtype=relative_coords.dtype)
+            relative_position_index[1:, 1:] = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
+            relative_position_index[0, 0:] = self.num_relative_distance - 3
+            relative_position_index[0:, 0] = self.num_relative_distance - 2
+            relative_position_index[0, 0] = self.num_relative_distance - 1
+
+            self.register_buffer("relative_position_index", relative_position_index)
+        else:
+            self.window_size = None
+            self.relative_position_bias_table = None
+            self.relative_position_index = None
+
+        self.attn_drop = nn.Dropout(attn_drop)
+        self.proj = nn.Linear(all_head_dim, dim)
+        self.proj_drop = nn.Dropout(proj_drop)
+
+    def forward(self, x, rel_pos_bias=None):
+        B, N, C = x.shape
+        qkv_bias = None
+        if self.q_bias is not None:
+            qkv_bias = torch.cat((self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias))
+        # qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
+        qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
+        qkv = qkv.reshape(B, N, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4)
+        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)
+
+        q = q * self.scale
+        attn = (q @ k.transpose(-2, -1))
+
+        if self.relative_position_bias_table is not None:
+            relative_position_bias = \
+                self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
+                    self.window_size[0] * self.window_size[1] + 1,
+                    self.window_size[0] * self.window_size[1] + 1, -1)  # Wh*Ww,Wh*Ww,nH
+            relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
+            attn = attn + relative_position_bias.unsqueeze(0)
+
+        if rel_pos_bias is not None:
+            attn = attn + rel_pos_bias
+
+        attn = attn.softmax(dim=-1)
+        attn = self.attn_drop(attn)
+
+        x = (attn @ v).transpose(1, 2).reshape(B, N, -1)
+        x = self.proj(x)
+        x = self.proj_drop(x)
+        return x
+
+
+class BlockEvaVit(nn.Module):
+    def __init__(self,
+                 dim,
+                 num_heads,
+                 mlp_ratio=4.,
+                 qkv_bias=False,
+                 qk_scale=None,
+                 drop=0.,
+                 attn_drop=0.,
+                 drop_path=0.,
+                 init_values=None,
+                 act_layer=nn.GELU,
+                 norm_layer=nn.LayerNorm,
+                 window_size=None,
+                 attn_head_dim=None):
+        super().__init__()
+        self.norm1 = norm_layer(dim)
+        self.attn = AttentionEvaVit(dim,
+                              num_heads=num_heads,
+                              qkv_bias=qkv_bias,
+                              qk_scale=qk_scale,
+                              attn_drop=attn_drop,
+                              proj_drop=drop,
+                              window_size=window_size,
+                              attn_head_dim=attn_head_dim)
+        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here
+        self.drop_path = DropPathEvaVit(drop_path) if drop_path > 0. else nn.Identity()
+        self.norm2 = norm_layer(dim)
+        mlp_hidden_dim = int(dim * mlp_ratio)
+        self.mlp = MlpEvaVit(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)
+
+        if init_values is not None and init_values > 0:
+            self.gamma_1 = nn.Parameter(init_values * torch.ones((dim)), requires_grad=True)
+            self.gamma_2 = nn.Parameter(init_values * torch.ones((dim)), requires_grad=True)
+        else:
+            self.gamma_1, self.gamma_2 = None, None
+
+    def forward(self, x, rel_pos_bias=None):
+        if self.gamma_1 is None:
+            x = x + self.drop_path(self.attn(self.norm1(x), rel_pos_bias=rel_pos_bias))
+            x = x + self.drop_path(self.mlp(self.norm2(x)))
+        else:
+            x = x + self.drop_path(self.gamma_1 * self.attn(self.norm1(x), rel_pos_bias=rel_pos_bias))
+            x = x + self.drop_path(self.gamma_2 * self.mlp(self.norm2(x)))
+        return x
+
+
+class PatchEmbedEvaVit(nn.Module):
+    """ Image to Patch Embedding
+    """
+    def __init__(self, img_size=224, patch_size=16, in_chans=3, embed_dim=768):
+        super().__init__()
+        img_size = to_2tuple(img_size)
+        patch_size = to_2tuple(patch_size)
+        num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0])
+        self.patch_shape = (img_size[0] // patch_size[0], img_size[1] // patch_size[1])
+        self.img_size = img_size
+        self.patch_size = patch_size
+        self.num_patches = num_patches
+
+        self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
+
+    def forward(self, x, **kwargs):
+        B, C, H, W = x.shape
+        # FIXME look at relaxing size constraints
+        assert H == self.img_size[0] and W == self.img_size[1], \
+            f"Input image size ({H}*{W}) doesn't match model ({self.img_size[0]}*{self.img_size[1]})."
+        x = self.proj(x).flatten(2).transpose(1, 2)
+        return x
+
+
+class RelativePositionBiasEvaVit(nn.Module):
+    def __init__(self, window_size, num_heads):
+        super().__init__()
+        self.window_size = window_size
+        self.num_relative_distance = (2 * window_size[0] - 1) * (2 * window_size[1] - 1) + 3
+        self.relative_position_bias_table = nn.Parameter(torch.zeros(self.num_relative_distance,
+                                                                     num_heads))  # 2*Wh-1 * 2*Ww-1, nH
+        # cls to token & token 2 cls & cls to cls
+
+        # get pair-wise relative position index for each token inside the window
+        coords_h = torch.arange(window_size[0])
+        coords_w = torch.arange(window_size[1])
+        coords = torch.stack(torch.meshgrid([coords_h, coords_w]))  # 2, Wh, Ww
+        coords_flatten = torch.flatten(coords, 1)  # 2, Wh*Ww
+        relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :]  # 2, Wh*Ww, Wh*Ww
+        relative_coords = relative_coords.permute(1, 2, 0).contiguous()  # Wh*Ww, Wh*Ww, 2
+        relative_coords[:, :, 0] += window_size[0] - 1  # shift to start from 0
+        relative_coords[:, :, 1] += window_size[1] - 1
+        relative_coords[:, :, 0] *= 2 * window_size[1] - 1
+        relative_position_index = \
+            torch.zeros(size=(window_size[0] * window_size[1] + 1,) * 2, dtype=relative_coords.dtype)
+        relative_position_index[1:, 1:] = relative_coords.sum(-1)  # Wh*Ww, Wh*Ww
+        relative_position_index[0, 0:] = self.num_relative_distance - 3
+        relative_position_index[0:, 0] = self.num_relative_distance - 2
+        relative_position_index[0, 0] = self.num_relative_distance - 1
+
+        self.register_buffer("relative_position_index", relative_position_index)
+
+        # trunc_normal_(self.relative_position_bias_table, std=.02)
+
+    def forward(self):
+        relative_position_bias = \
+            self.relative_position_bias_table[self.relative_position_index.view(-1)].view(
+                self.window_size[0] * self.window_size[1] + 1,
+                self.window_size[0] * self.window_size[1] + 1, -1)  # Wh*Ww,Wh*Ww,nH
+        return relative_position_bias.permute(2, 0, 1).contiguous()  # nH, Wh*Ww, Wh*Ww
+
+
+class VisionTransformerEvaVit(nn.Module):
+    """ Vision Transformer with support for patch or hybrid CNN input stage
+    """
+    def __init__(self,
+                 img_size=224,
+                 patch_size=16,
+                 in_chans=3,
+                 num_classes=1000,
+                 embed_dim=768,
+                 depth=12,
+                 num_heads=12,
+                 mlp_ratio=4.,
+                 qkv_bias=False,
+                 qk_scale=None,
+                 drop_rate=0.,
+                 attn_drop_rate=0.,
+                 drop_path_rate=0.,
+                 norm_layer=nn.LayerNorm,
+                 init_values=None,
+                 use_abs_pos_emb=True,
+                 use_rel_pos_bias=False,
+                 use_shared_rel_pos_bias=False,
+                 use_mean_pooling=True,
+                 init_scale=0.001,
+                 use_checkpoint=False):
+        super().__init__()
+        self.image_size = img_size
+        self.num_classes = num_classes
+        self.num_features = self.embed_dim = embed_dim  # num_features for consistency with other models
+
+        self.patch_embed = PatchEmbedEvaVit(img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim)
+        num_patches = self.patch_embed.num_patches
+
+        self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim))
+        if use_abs_pos_emb:
+            self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim))
+        else:
+            self.pos_embed = None
+        self.pos_drop = nn.Dropout(p=drop_rate)
+
+        if use_shared_rel_pos_bias:
+            self.rel_pos_bias = RelativePositionBiasEvaVit(window_size=self.patch_embed.patch_shape, num_heads=num_heads)
+        else:
+            self.rel_pos_bias = None
+        self.use_checkpoint = use_checkpoint
+
+        dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)]  # stochastic depth decay rule
+        self.use_rel_pos_bias = use_rel_pos_bias
+        self.blocks = nn.ModuleList([
+            BlockEvaVit(dim=embed_dim,
+                  num_heads=num_heads,
+                  mlp_ratio=mlp_ratio,
+                  qkv_bias=qkv_bias,
+                  qk_scale=qk_scale,
+                  drop=drop_rate,
+                  attn_drop=attn_drop_rate,
+                  drop_path=dpr[i],
+                  norm_layer=norm_layer,
+                  init_values=init_values,
+                  window_size=self.patch_embed.patch_shape if use_rel_pos_bias else None) for i in range(depth)
+        ])
+        #         self.norm = nn.Identity() if use_mean_pooling else norm_layer(embed_dim)
+        #         self.fc_norm = norm_layer(embed_dim) if use_mean_pooling else None
+        #         self.head = nn.Linear(embed_dim, num_classes) if num_classes > 0 else nn.Identity()
+
+        if self.pos_embed is not None:
+            trunc_normal_(self.pos_embed, std=.02)
+        trunc_normal_(self.cls_token, std=.02)
+        # trunc_normal_(self.mask_token, std=.02)
+        #         if isinstance(self.head, nn.Linear):
+        #             trunc_normal_(self.head.weight, std=.02)
+        self.apply(self._init_weights)
+        self.fix_init_weight()
+        self.ln_vision =  nn.LayerNorm(self.num_features)
+
+    def fix_init_weight(self):
+        def rescale(param, layer_id):
+            param.div_(math.sqrt(2.0 * layer_id))
+
+        for layer_id, layer in enumerate(self.blocks):
+            rescale(layer.attn.proj.weight.data, layer_id + 1)
+            rescale(layer.mlp.fc2.weight.data, layer_id + 1)
+
+    def _init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            trunc_normal_(m.weight, std=.02)
+            if isinstance(m, nn.Linear) and m.bias is not None:
+                nn.init.constant_(m.bias, 0)
+        elif isinstance(m, nn.LayerNorm):
+            nn.init.constant_(m.bias, 0)
+            nn.init.constant_(m.weight, 1.0)
+
+    _initialize_weights = _init_weights
+    
+    def get_classifier(self):
+        return self.head
+
+    def reset_classifier(self, num_classes, global_pool=''):
+        self.num_classes = num_classes
+        self.head = nn.Linear(self.embed_dim, num_classes) if num_classes > 0 else nn.Identity()
+
+    def forward_features(self, x):
+        x = self.patch_embed(x)
+        batch_size, seq_len, _ = x.size()
+
+        cls_tokens = self.cls_token.expand(batch_size, -1, -1)  # stole cls_tokens impl from Phil Wang, thanks
+        x = torch.cat((cls_tokens, x), dim=1)
+        if self.pos_embed is not None:
+            x = x + self.pos_embed
+        x = self.pos_drop(x)
+
+        rel_pos_bias = self.rel_pos_bias() if self.rel_pos_bias is not None else None
+        for blk in self.blocks:
+            if self.use_checkpoint:
+                x = checkpoint.checkpoint(blk, x, rel_pos_bias)
+            else:
+                x = blk(x, rel_pos_bias)
+        return x
+
+    def forward(self, x):
+        x = self.forward_features(x)
+        #         x = self.head(x)
+        return x
+
+    def get_intermediate_layers(self, x):
+        x = self.patch_embed(x)
+        batch_size, seq_len, _ = x.size()
+
+        cls_tokens = self.cls_token.expand(batch_size, -1, -1)  # stole cls_tokens impl from Phil Wang, thanks
+        x = torch.cat((cls_tokens, x), dim=1)
+        if self.pos_embed is not None:
+            x = x + self.pos_embed
+        x = self.pos_drop(x)
+
+        features = []
+        rel_pos_bias = self.rel_pos_bias() if self.rel_pos_bias is not None else None
+        for blk in self.blocks:
+            x = blk(x, rel_pos_bias)
+            features.append(x)
+
+        return features
+
+    def get_num_layer(self, var_name=""):
+        if var_name in ("cls_token", "mask_token", "pos_embed"):
+            return 0
+        elif var_name.startswith("patch_embed"):
+            return 0
+        elif var_name.startswith("rel_pos_bias"):
+            return len(self.blocks) - 1
+        elif var_name.startswith("blocks"):
+            layer_id = int(var_name.split('.')[1])
+            return layer_id + 1
+        else:
+            return len(self.blocks)
+
+
+def create_eva_vit_g(img_size=224, drop_path_rate=0.4, use_checkpoint=False, precision="fp16",         cache_dir="./",):
+    model = VisionTransformerEvaVit(
+        img_size=img_size,
+        patch_size=14,
+        use_mean_pooling=False,
+        embed_dim=1408,
+        depth=39,
+        num_heads=1408 // 88,
+        mlp_ratio=4.3637,
+        qkv_bias=True,
+        drop_path_rate=drop_path_rate,
+        norm_layer=partial(nn.LayerNorm, eps=1e-6),
+        use_checkpoint=use_checkpoint,
+    )
+    cache_path = cache_dir
+    state_dict = torch.load(cache_path+"/eva_vit_g.pth", map_location="cpu")    
+    interpolate_pos_embed(model, state_dict)
+
+    incompatible_keys = model.load_state_dict(state_dict, strict=False)
+    print(incompatible_keys)
+
+    return model
 
 class BertEmbeddings(nn.Module):
     """Construct the embeddings from word and position embeddings."""
@@ -1491,10 +1906,6 @@ class VisionTransformer(nn.Module):
 
         return x
 
-    @torch.jit.ignore()
-    def load_pretrained(self, checkpoint_path, prefix=""):
-        _load_weights(self, checkpoint_path, prefix)
-
 
 @torch.no_grad()
 def _load_weights(model: VisionTransformer, checkpoint_path: str, prefix: str = ""):
@@ -1642,19 +2053,9 @@ class Blip2Base(PreTrainedModel):
         tokenizer.add_special_tokens({"bos_token": "[DEC]"})
         return tokenizer
 
-    def maybe_autocast(self, dtype=torch.float16):
-        # if on cpu, don't use autocast
-        # if on gpu, use autocast with dtype if provided, otherwise use torch.float16
-        enable_autocast = self.device != torch.device("cpu")
-
-        if enable_autocast:
-            return torch.cuda.amp.autocast(dtype=dtype)
-        else:
-            return contextlib.nullcontext()
-
     @classmethod
     def init_Qformer(cls, encoder_config, num_query_token, vision_width, cross_attention_freq=2, cache_dir=""):
-        print ("loading")
+        #print ("loading")
         encoder_config = BertConfig.from_pretrained("bert-base-uncased")
         encoder_config.encoder_width = vision_width
         # insert cross-attention layer every other block
@@ -1666,74 +2067,6 @@ class Blip2Base(PreTrainedModel):
         query_tokens.data.normal_(mean=0.0, std=encoder_config.initializer_range)
         return Qformer, query_tokens
 
-    def load_from_pretrained(self, url_or_filename):
-        if is_url(url_or_filename):
-            cached_file = download_cached_file(url_or_filename, check_hash=False, progress=True)
-            checkpoint = torch.load(cached_file, map_location="cpu")
-        elif os.path.isfile(url_or_filename):
-            checkpoint = torch.load(url_or_filename, map_location="cpu")
-        else:
-            raise RuntimeError("checkpoint url or path is invalid")
-
-        state_dict = checkpoint["model"]
-
-        msg = self.load_state_dict(state_dict, strict=False)
-
-        # logging.info("Missing keys {}".format(msg.missing_keys))
-        logging.info("load checkpoint from %s" % url_or_filename)
-
-        return msg
-
-    def _lemmatize(self, answers):
-        def apply(answer):
-            doc = self.lemmatizer(answer)
-
-            words = []
-            for token in doc:
-                if token.pos_ in ["NOUN", "VERB"]:
-                    words.append(token.lemma_)
-                else:
-                    words.append(token.text)
-            answer = " ".join(words)
-
-            return answer
-
-        return [apply(answer) for answer in answers]
-
-    @property
-    def lemmatizer(self):
-        if self._lemmatizer is None:
-            try:
-                import spacy
-
-                self._lemmatizer = spacy.load("en_core_web_sm")
-            except ImportError:
-                logging.error("""
-                    Please install spacy and en_core_web_sm model to apply lemmatization.
-                    python -m spacy download en_core_web_sm
-                    OR
-                    import spacy.cli
-                    spacy.cli.download("en_core_web_sm")
-                    """)
-                exit(1)
-
-        return self._lemmatizer
-
-
-def disabled_train(self, mode=True):
-    """Overwrite model.train with this function to make sure train/eval mode
-    does not change anymore."""
-    return self
-
-
-class LayerNorm(nn.LayerNorm):
-    """Subclass torch's LayerNorm to handle fp16."""
-    def forward(self, x: torch.Tensor):
-        orig_type = x.dtype
-        ret = super().forward(x.type(torch.float32))
-        return ret.type(orig_type)
-
-
 
 
 class VectorQuantizer2(nn.Module):
@@ -1977,43 +2310,28 @@ class Blip2QformerQuantizer(Blip2Base):
         )
         self.distill_image_proj = nn.Linear(num_query_token * 32, image_features_dim)
 
-    @classmethod
-    def load_from_pretrained(cls, config, pretrained_model_path, **kwargs):
-        img_size = kwargs.get("image_size", 224)
-        num_query_token = kwargs.get("num_query_token", 32)
-        cross_attention_freq = kwargs.get("cross_attention_freq", 2)
-
-        drop_path_rate = kwargs.get("drop_path_rate", 0)
-        use_grad_checkpoint = kwargs.get("use_grad_checkpoint", False)
-        freeze_vit = kwargs.get("freeze_vit", True)
-        cache_dir = kwargs.get("cache_dir", "./")        
-
-        max_txt_len = kwargs.get("max_txt_len", 32)
-
-        model = cls(config,
-            img_size=img_size,
-            drop_path_rate=drop_path_rate,
-            use_grad_checkpoint=use_grad_checkpoint,
-            freeze_vit=freeze_vit,
-            num_query_token=num_query_token,
-            cross_attention_freq=cross_attention_freq,
-            max_txt_len=max_txt_len,
-            cache_dir=cache_dir,
-        )
-
-        ckpt = torch.load(cache_dir+pretrained_model_path, map_location="cpu")
-        missing, unexcepted = model.load_state_dict(ckpt, strict=False)
-        #print('**** missing keys: ', missing)
-        #print('***unexpected keys:', unexcepted)
-        return model
+    def get_codebook_indices(self, visual_encoder, image):
+        with torch.no_grad():
+            image_embeds = visual_encoder.ln_vision(visual_encoder(image))
+            image_atts = torch.ones(image_embeds.size()[:-1], dtype=torch.long).to(image.device)
+            query_tokens = self.query_tokens.expand(image_embeds.shape[0], -1, -1)
+            query_output = self.Qformer.bert(
+                query_embeds=query_tokens,
+                encoder_hidden_states=image_embeds,
+                encoder_attention_mask=image_atts,
+                return_dict=True,
+            )
 
+            query_output_down = self.encode_task_layer(query_output.last_hidden_state)
+            quant, loss_embed, embed_ind = self.quantize(query_output_down)
+            embed_ind = embed_ind.reshape(quant.shape[0], -1)
 
+        return embed_ind
 
     def get_codebook_indices(self, visual_encoder, image):
         with torch.no_grad():
-            with self.maybe_autocast():
-                image_embeds = visual_encoder.ln_vision(visual_encoder(image))
-                image_atts = torch.ones(image_embeds.size()[:-1], dtype=torch.long).to(image.device)
+            image_embeds = visual_encoder.ln_vision(visual_encoder(image))
+            image_atts = torch.ones(image_embeds.size()[:-1], dtype=torch.long).to(image.device)
             query_tokens = self.query_tokens.expand(image_embeds.shape[0], -1, -1)
             query_output = self.Qformer.bert(
                 query_embeds=query_tokens,
@@ -2029,7 +2347,7 @@ class Blip2QformerQuantizer(Blip2Base):
             query_output_up = self.decode_task_layer(quant)
 
         return embed_ind, query_output_up
-
+    
     def get_codebook_entry(self, indices):
         with torch.no_grad():
             quant_embedding = self.quantize.get_codebook_entry(indices)
@@ -2058,12 +2376,10 @@ class Blip2QformerQuantizer(Blip2Base):
                            precision="fp32",
                            cache_dir="./"):
         visual_encoder = create_eva_vit_g(img_size, drop_path_rate, use_grad_checkpoint, precision, cache_dir=cache_dir)
-        visual_encoder.ln_vision = LayerNorm(visual_encoder.num_features)
+        visual_encoder.ln_vision = nn.LayerNorm(visual_encoder.num_features)
         for name, param in visual_encoder.named_parameters():
             param.requires_grad = False
         visual_encoder = visual_encoder.eval()
-        visual_encoder.train = disabled_train
-        logging.info("freeze vision encoder")
         visual_encoder.ln_vision.weight.requires_grad = False
         visual_encoder.ln_vision.bias.requires_grad = False
         return visual_encoder
@@ -2094,12 +2410,10 @@ class Seed2Tokenizer(PreTrainedModel):
         self.register_buffer("latents",torch.randn(shape_latents, generator=None, layout=torch.strided))
         
 
-        shape_noise = torch.Size([1, 1024])
-        self.register_buffer("noise",torch.randn(shape_noise, generator=None, layout=torch.strided))
-
+        
         self.model = model
         self.processor = processor
-        self.visual_encoder = VisionTransformerEvaClip(
+        self.visual_encoder = VisionTransformerEvaVit(
             img_size=image_size,
             patch_size=14,
             use_mean_pooling=False,
@@ -2128,10 +2442,13 @@ class Seed2Tokenizer(PreTrainedModel):
         if len(image_torch.shape) == 3:
             image_torch = image_torch.unsqueeze(0)
 
+        image_torch = image_torch.to(dtype=self.latents.dtype)
+        image_torch = image_torch.to(self.device)
         # img = image_torch.to(self.device)
         img = image_torch
         #if self.fp16:
         #    img = img.half()
+        print (img.dtype)
         with torch.no_grad():
             id, _ = self.model.get_codebook_indices(visual_encoder, img)
         return id.view(img.shape[0], -1)
@@ -2146,7 +2463,6 @@ class Seed2Tokenizer(PreTrainedModel):
 
         image = diffusion_model(
             image_embeds=image_embeds,
-            negative_image_embeds=negative_image_embeds,
             guidance_scale=guidance_scale,
             noise_level=noise_level,
             num_inference_steps=num_inference_steps,
@@ -2178,5 +2494,6 @@ class Seed2Tokenizer(PreTrainedModel):
             image_torch = self.processor(image_pil)
 
             image_torch = image_torch.to(self.device)
+        image_torch = image_torch.to(dtype=self.latents.dtype)
         return self.encode(image_torch, visual_encoder)