[Init] models

model/blip2_pretrained_flant5xxl.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b3839ea6c617f315ead9bf4036bbb0f0cf6bf62695ecfc14968ea626af03a29
+size 433481467

model/videochat_13b.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd78ad4b1fff617ebd603349d9409495b88938135338576e0c76f6268c6f20d0
+size 825484451

models/eva_vit.py

@@ -16,6 +16,9 @@ import torch.nn as nn
 import torch.nn.functional as F
 import torch.utils.checkpoint as checkpoint
 from timm.models.layers import drop_path, to_2tuple, trunc_normal_
+from timm.models.registry import register_model
+
+from utils.misc import download_cached_file
 
 
 def _cfg(url='', **kwargs):
@@ -588,7 +591,10 @@ def create_eva_vit_g(
         gmhra_dropout=gmhra_dropout, 
     )  
     if vit_model_path is not None and os.path.isfile(vit_model_path):
-        state_dict = torch.load(vit_model_path, map_location="cpu")    
+        cached_file = download_cached_file(
+            vit_model_path, check_hash=False, progress=True
+        )
+        state_dict = torch.load(cached_file, map_location="cpu")    
         print(f"Load ViT model from: {vit_model_path}")
         interpolate_pos_embed(model, state_dict)
         msg = load_state_dict(model, state_dict, strict=False)