Model weight access updates

brainscore_vision/models/bp_resnet50_julios/metadata.yaml

@@ -0,0 +1,14 @@
+models:
+    bp_resnet50_julios:
+        architecture: DCNN
+        model_family: resnet
+        total_parameter_count: 25503912
+        trainable_parameter_count: 25503912
+        total_layers: 151
+        trainable_layers: 54
+        model_size_MB: 102.02
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/bp_resnet50_julios
+        huggingface_link: null
+        extra_notes: null

brainscore_vision/models/bp_resnet50_julios/model.py

@@ -25,13 +25,13 @@ def get_model(name):
     assert name == 'bp_resnet50_julios'
     ckpt_name = 'bp_export.pt'
     download_weights(
-        bucket='brainscore-vision',
-        folder_path='models/bp_resnet50_julios',
+        bucket='brainscore-storage',
+        folder_path='brainscore-vision/models/bp_resnet50_julios',
         filename_version_sha=[(ckpt_name, 'PAH2BMGhcw9tWUH6RwTIADHodw.fyG4G', 'e504fc533aaec453f24e73a86e47ca6df3394c53')],
         save_directory=Path(__file__).parent
     )
     ckpt_path = Path(__file__).parent / ckpt_name
-    model = torch.load(ckpt_path)
+    model = torch.load(ckpt_path, map_location=torch.device('cpu'))
     preprocessing = functools.partial(load_preprocess_images, image_size=224)
     print(preprocessing)
     wrapper = PytorchWrapper(identifier='bp_resnet50_julios', model=model, preprocessing=preprocessing)

brainscore_vision/models/dbp_resnet50_julios/metadata.yaml

@@ -0,0 +1,14 @@
+models:
+    dbp_resnet50_julios:
+        architecture: DCNN
+        model_family: resnet
+        total_parameter_count: 25503912
+        trainable_parameter_count: 25503912
+        total_layers: 151
+        trainable_layers: 54
+        model_size_MB: 102.02
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/dbp_resnet50_julios
+        huggingface_link: null
+        extra_notes: null

brainscore_vision/models/dbp_resnet50_julios/model.py

@@ -25,13 +25,13 @@ def get_model(name):
     assert name == 'dbp_resnet50_julios'
     ckpt_name = 'dbp_export.pt'
     download_weights(
-        bucket='brainscore-vision', 
-        folder_path='models/dbp_resnet50_julios',
+        bucket='brainscore-storage',
+        folder_path='brainscore-vision/models/dbp_resnet50_julios',
         filename_version_sha=[(ckpt_name, 'HpyXY.W.WtS8zD.wJ7BVfvSbQDX.1Md4', '6efa18bfe4b70221d7e88ac2d2279d08042f23da')],
         save_directory=Path(__file__).parent
     )
     ckpt_path = Path(__file__).parent / ckpt_name
-    model = torch.load(ckpt_path)    
+    model = torch.load(ckpt_path, map_location=torch.device('cpu'))
     preprocessing = functools.partial(load_preprocess_images, image_size=224)
     print(preprocessing)
     wrapper = PytorchWrapper(identifier='dbp_resnet50_julios', model=model, preprocessing=preprocessing)

brainscore_vision/models/temporal_model_AVID_CMA/metadata.yaml

@@ -0,0 +1,53 @@
+models:
+    AVID-Audioset:
+        architecture: DCNN
+        model_family: null
+        total_parameter_count: 15379136
+        trainable_parameter_count: 15379136
+        total_layers: 94
+        trainable_layers: 36
+        model_size_MB: 61.52
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_AVID_CMA
+        huggingface_link: null
+        extra_notes: null
+    AVID-CMA-Audioset:
+        architecture: DCNN
+        model_family: null
+        total_parameter_count: 15379136
+        trainable_parameter_count: 15379136
+        total_layers: 94
+        trainable_layers: 36
+        model_size_MB: 61.52
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_AVID_CMA
+        huggingface_link: null
+        extra_notes: null
+    AVID-CMA-Kinetics400:
+        architecture: DCNN
+        model_family: null
+        total_parameter_count: 15379136
+        trainable_parameter_count: 15379136
+        total_layers: 94
+        trainable_layers: 36
+        model_size_MB: 61.52
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_AVID_CMA
+        huggingface_link: null
+        extra_notes: null
+    AVID-Kinetics400:
+        architecture: DCNN
+        model_family: null
+        total_parameter_count: 15379136
+        trainable_parameter_count: 15379136
+        total_layers: 94
+        trainable_layers: 36
+        model_size_MB: 61.52
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_AVID_CMA
+        huggingface_link: null
+        extra_notes: null

brainscore_vision/models/temporal_model_AVID_CMA/model.py

@@ -19,36 +19,36 @@ def get_model(identifier):
     if identifier == 'AVID-CMA-Kinetics400':
         cfg_path = os.path.join(HOME, "configs/main/avid-cma/kinetics/InstX-N1024-PosW-N64-Top32.yaml")
         weight_path = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path="temporal_model_AVID-CMA/AVID-CMA_Kinetics_InstX-N1024-PosW-N64-Top32_checkpoint.pth.tar",
-            version_id="yx9Pbq3SuNOOd4sX7csTolaHD1iTCx8y",
+            version_id="null",
             sha1="6efe4464ca654a56affff766acf24e89e6f3ffbf"
         )
 
     elif identifier == 'AVID-CMA-Audioset':
         cfg_path = os.path.join(HOME, "configs/main/avid-cma/audioset/InstX-N1024-PosW-N64-Top32.yaml")
         weight_path = load_weight_file(
-            bucket="brainscore-vision",
-            relative_path="temporal_model_AVID_CMA/AVID-CMA_Audioset_InstX-N1024-PosW-N64-Top32_checkpoint.pth.tar",
-            version_id="jSaZgbUohM0ZeoEUUKZiLBo6iz_v8VvQ",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
+            relative_path="temporal_model_AVID-CMA/AVID-CMA_Audioset_InstX-N1024-PosW-N64-Top32_checkpoint.pth.tar",
+            version_id="null",
             sha1="9db5eba9aab6bdbb74025be57ab532df808fe3f6"
         )
 
     elif identifier == 'AVID-Kinetics400':
         cfg_path = os.path.join(HOME, "configs/main/avid/kinetics/Cross-N1024.yaml")
         weight_path = load_weight_file(
-            bucket="brainscore-vision",
-            relative_path="temporal_model_AVID_CMA/AVID_Kinetics_Cross-N1024_checkpoint.pth.tar",
-            version_id="XyKt0UOUFsuuyrl6ZREivK8FadRPx34u",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
+            relative_path="temporal_model_AVID-CMA/AVID_Kinetics_Cross-N1024_checkpoint.pth.tar",
+            version_id="null",
             sha1="d3a04f856d29421ba8de37808593a3fad4d4794f"
         )
 
     elif identifier == 'AVID-Audioset':
         cfg_path = os.path.join(HOME, "configs/main/avid/audioset/Cross-N1024.yaml")
         weight_path = load_weight_file(
-            bucket="brainscore-vision",
-            relative_path="temporal_model_AVID_CMA/AVID_Audioset_Cross-N1024_checkpoint.pth.tar",
-            version_id="0Sxuhn8LsYXQC4FnPfJ7rw7uU6kDlKgc",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
+            relative_path="temporal_model_AVID-CMA/AVID_Audioset_Cross-N1024_checkpoint.pth.tar",
+            version_id="null",
             sha1="b48d8428a1a2526ccca070f810333df18bfce5fd"
         )
 

brainscore_vision/models/temporal_model_GDT/metadata.yaml

@@ -0,0 +1,40 @@
+models:
+    GDT-HowTo100M:
+        architecture: DCNN
+        model_family: null
+        total_parameter_count: 31300125
+        trainable_parameter_count: 31300125
+        total_layers: 160
+        trainable_layers: 37
+        model_size_MB: 125.2
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_GDT
+        huggingface_link: null
+        extra_notes: null
+    GDT-IG65M:
+        architecture: DCNN
+        model_family: null
+        total_parameter_count: 31300125
+        trainable_parameter_count: 31300125
+        total_layers: 160
+        trainable_layers: 37
+        model_size_MB: 125.2
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_GDT
+        huggingface_link: null
+        extra_notes: null
+    GDT-Kinetics400:
+        architecture: DCNN
+        model_family: null
+        total_parameter_count: 31300125
+        trainable_parameter_count: 31300125
+        total_layers: 160
+        trainable_layers: 37
+        model_size_MB: 125.2
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_GDT
+        huggingface_link: null
+        extra_notes: null

brainscore_vision/models/temporal_model_GDT/model.py

@@ -20,23 +20,23 @@ def get_model(identifier):
 
     if dataset == "Kinetics400":
         pth = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path="temporal_model_GDT/gdt_K400.pth",
-            version_id="JpU_tnCzrbTejn6sOrQMk8eRsJ97yFgt",
+            version_id="null",
             sha1="7f12c60670346b1aab15194eb44c341906e1bca6"
         )
     elif dataset == "IG65M":
         pth = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path="temporal_model_GDT/gdt_IG65M.pth",
-            version_id="R.NoD6VAbFbJdf8tg5jnXIWB3hQ8GlSD",
+            version_id="null",
             sha1="3dcee3af61691e1e7e47e4b115be6808f4ea8172"
         )
     elif dataset == "HowTo100M":
         pth = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path="temporal_model_GDT/gdt_HT100M.pth",
-            version_id="BVRl9t_134PoKZCn9W54cyfkImCW2ioq",
+            version_id="null",
             sha1="a9a979c82e83b955794814923af736eb34e6f080"
         )
     else:

brainscore_vision/models/temporal_model_S3D_text_video/metadata.yaml

@@ -0,0 +1,14 @@
+models:
+    s3d-HowTo100M:
+        architecture: DCNN
+        model_family: s3d
+        total_parameter_count: 31211336
+        trainable_parameter_count: 31211336
+        total_layers: 376
+        trainable_layers: 117
+        model_size_MB: 124.85
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_S3D_text_video
+        huggingface_link: null
+        extra_notes: null

brainscore_vision/models/temporal_model_S3D_text_video/model.py

@@ -39,16 +39,16 @@ def get_model(identifier="s3d-HowTo100M"):
     model_name = identifier
 
     model_pth = load_weight_file(
-        bucket="brainscore-vision",
+        bucket="brainscore-storage", folder_name="brainscore-vision/models",
         relative_path="temporal_model_S3D_text_video/s3d_howto100m.pth",
-        version_id="hRp6I8bpwreIMUVL0H.zCdK0hqRggL7n",
+        version_id="null",
         sha1="31e99d2a1cd48f2259ca75e719ac82c8b751ea75"
     )
 
     dict_pth = load_weight_file(
-        bucket="brainscore-vision",
+        bucket="brainscore-storage", folder_name="brainscore-vision/models",
         relative_path="temporal_model_S3D_text_video/s3d_dict.npy",
-        version_id="4NxVLe8DSL6Uue0F7e2rz8HZuOk.tkBI",
+        version_id="null",
         sha1="d368ff7d397ec8240f1f963b5efe8ff245bac35f"
     )
 

brainscore_vision/models/temporal_model_VideoMAE/metadata.yaml

@@ -0,0 +1,27 @@
+models:
+    VideoMAE-V1-B:
+        architecture: DCNN, Transformer
+        model_family: null
+        total_parameter_count: 101307264
+        trainable_parameter_count: 101307264
+        total_layers: 292
+        trainable_layers: 84
+        model_size_MB: 405.23
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_VideoMAE
+        huggingface_link: null
+        extra_notes: null
+    VideoMAE-V1-L:
+        architecture: DCNN, Transformer
+        model_family: null
+        total_parameter_count: 330389504
+        trainable_parameter_count: 330389504
+        total_layers: 460
+        trainable_layers: 132
+        model_size_MB: 1321.56
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_VideoMAE
+        huggingface_link: null
+        extra_notes: null

brainscore_vision/models/temporal_model_VideoMAE/model.py

@@ -37,18 +37,18 @@ def get_model(identifier, num_frames=16):
     if identifier == "VideoMAE-V1-B":
         model_name = "pretrain_videomae_base_patch16_224"
         pth = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path='temporal_model_VideoMAE/vit_b_k400_pt_1200e.pth',
-            version_id="Oi3VboRZujNyZAcwf5q7XZ2M8q1cPO6o",
+            version_id="null",
             sha1="8faf42df13f619a8970d653695e98f0643760b54"
         )
         num_blocks = 12
     elif identifier == "VideoMAE-V1-L":
         model_name = "pretrain_videomae_large_patch16_224"
         pth = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path='temporal_model_VideoMAE/vit_l_k400_pt_1200e.pth',
-            version_id="MiPfczDaVponDGuUBrEPqmT.BiVvh_I1",
+            version_id="null",
             sha1="7ff6acbba221f85d7148223ec932ad7c57f2f40c"
         )
         num_blocks = 24

brainscore_vision/models/temporal_model_VideoMAEv2/metadata.yaml

@@ -0,0 +1,27 @@
+models:
+    VideoMAE-V2-B:
+        architecture: DCNN, Transformer
+        model_family: null
+        total_parameter_count: 86996200
+        trainable_parameter_count: 86996200
+        total_layers: 177
+        trainable_layers: 50
+        model_size_MB: 347.98
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_VideoMAEv2
+        huggingface_link: null
+        extra_notes: null
+    VideoMAE-V2-G:
+        architecture: DCNN, Transformer
+        model_family: null
+        total_parameter_count: 1013019752
+        trainable_parameter_count: 1013019752
+        total_layers: 569
+        trainable_layers: 162
+        model_size_MB: 4052.08
+        training_dataset: null
+        task_specialization: null
+        brainscore_link: https://github.com/brain-score/vision/tree/master/brainscore_vision/models/temporal_model_VideoMAEv2
+        huggingface_link: null
+        extra_notes: null

brainscore_vision/models/temporal_model_VideoMAEv2/model.py

@@ -51,19 +51,19 @@ def get_model(identifier):
     if identifier == "VideoMAE-V2-G":
         model_name = "vit_giant_patch14_224"
         pth = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path="temporal_model_VideoMAEv2/vit_g_hybrid_pt_1200e.pth",
-            version_id="TxtkfbeMV105dzpzTwi0Kn5glnvQvIrq",
+            version_id="null",
             sha1="9048f2bc0b0c7ba4d0e5228f3a7c0bef4dbaca69"
         )
         num_blocks = 40
         feature_map_size = 16
     elif identifier == "VideoMAE-V2-B":
         model_name = "vit_base_patch16_224"
         pth = load_weight_file(
-            bucket="brainscore-vision",
+            bucket="brainscore-storage", folder_name="brainscore-vision/models",
             relative_path="temporal_model_VideoMAEv2/vit_b_hybrid_pt_800e.pth",
-            version_id="rRjpYq21dAQ5KaCLbEHK.YaLZ_fbMPKw",
+            version_id="null",
             sha1="1e3602691964b1eb6f7c33529119243a5b235635"
         )
         num_blocks = 12