Sync with C-RADIOv2-VLM-H-RC3 (#2)

- Sync with nvidia/C-RADIOv2-VLM-H-RC3 (a1fc2a754c7ce39bee9f0f1ebdf1ea32dc0f5ff8)
- Update README for release (581f0a8268fb26143fe249849e3f61b53c33a695)
- Remove custom load_state_dict (fe079c89a3bc7b68fb3ede1163d3c90c8679fef9)

README.md

@@ -6,6 +6,8 @@ license_link: https://developer.download.nvidia.com/licenses/nvidia-open-model-l
 
 # Model Overview
 
+[[**Github**](https://github.com/NVlabs/RADIO)]  [[**CVPR 2025**](https://arxiv.org/abs/2412.07679)]  [[**CVPR 2024**](https://arxiv.org/abs/2312.06709)]
+
 ## Description
 
 This model performs visual feature extraction.
@@ -78,7 +80,7 @@ import torch
 from PIL import Image
 from transformers import AutoModel, CLIPImageProcessor
 
-hf_repo = "nvidia/C-RADIOv2-H"
+hf_repo = "nvidia/C-RADIOv2-VLM-H"
 
 image_processor = CLIPImageProcessor.from_pretrained(hf_repo)
 model = AutoModel.from_pretrained(hf_repo, trust_remote_code=True)
@@ -224,4 +226,4 @@ Field                                               |  Response
 Model Application(s):                               |  Generation of visual embeddings
 Describe the life critical impact (if present).   |  Not Applicable
 Use Case Restrictions:                              |  Abide by NVIDIA Open Model License Agreement
-Model and dataset restrictions:            |  The Principle of least privilege (PoLP) is applied limiting access for dataset generation and model development. Restrictions enforce dataset access during training, and dataset license constraints adhered to.
+Model and dataset restrictions:            |  The Principle of least privilege (PoLP) is applied limiting access for dataset generation and model development. Restrictions enforce dataset access during training, and dataset license constraints adhered to.

config.json

@@ -16,7 +16,7 @@
     "cache_dir": null,
     "channels_last": false,
     "checkpoint_hist": 10,
-    "chk_keep_forever": 50,
+    "chk_keep_forever": 100,
     "class_map": "",
     "clip_grad": null,
     "clip_mode": "norm",
@@ -31,6 +31,7 @@
     "crop_pct": null,
     "cutmix": 0.0,
     "cutmix_minmax": null,
+    "damp": null,
     "dataset_download": false,
     "debug_full_knn": false,
     "decay_epochs": 90,
@@ -64,7 +65,7 @@
     "force_new_wandb_id": false,
     "force_spectral_reparam": true,
     "freeze_bn": false,
-    "fsdp": true,
+    "fsdp": false,
     "full_equivariance": false,
     "fuser": "",
     "gp": null,
@@ -169,6 +170,15 @@
         "name": "siglip2-g",
         "type": "siglip2",
         "use_summary": true
+      },
+      {
+        "fd_normalize": false,
+        "feature_distillation": true,
+        "input_size": 384,
+        "model": "siglip2-g-384",
+        "name": "siglip2-g-dirty",
+        "type": "siglip2",
+        "use_summary": false
       }
     ],
     "torchcompile": null,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e00385571b42a060742e3dd14ea7ac17f099968482115dc49a66aac34d1aa0a2
+oid sha256:96ff3bfec4f732d68a0c38c41a49de043abd2503df24481526ea87d26dd6a4f5
 size 2606616120