Replace checkpoint with submitted RE10K-only 24-view model

README.md

@@ -9,19 +9,19 @@ tags:
 
 # RLSplat
 
-This model repository contains inference-only RLSplat checkpoints in Safetensors format.
+This model repository contains the inference-only RLSplat checkpoint in Safetensors format.
 
-| Checkpoint | Training data | Default context views | Folder |
+| Checkpoint | Training data | Training context views | Folder |
 | --- | --- | ---: | --- |
-| RLSplat | RE10K and DL3DV | dynamic 2–24 | `re10k-dl3dv-dyn2-24view` |
+| RLSplat | RE10K | dynamic 2–24 (maximum 24) | `re10k-24view` |
 
-The checkpoint was exported from the 5,000-step `gspo5k_final` Lightning checkpoint. Optimizer state, training metadata, and training-only tensors are not included.
+The checkpoint was exported from the submitted 5,000-step RE10K Lightning checkpoint. Optimizer state, training metadata, and training-only tensors are not included.
 
 Use the model with the [RLSplat evaluation code](https://github.com/K-nowing/RLSplat/tree/release/evaluation-only-iclr26):
 
 ```bash
 python -m src.main '+evaluation=re10k' \
-  checkpointing.load=hf://Knowing/RLSplat/re10k-dl3dv-dyn2-24view/model.safetensors \
+  checkpointing.load=hf://Knowing/RLSplat/re10k-24view/model.safetensors \
   'dataset.re10k.roots=[/path/to/re10k]' \
   dataset.re10k.view_sampler.index_path=assets/evaluation_index_re10k.json
 ```

re10k-24view/config.json

@@ -0,0 +1,10 @@
+{
+  "model_type": "rlsplat",
+  "variant": "re10k-24view",
+  "training_data": "RE10K",
+  "context_views": "dynamic 2\u201324; maximum 24",
+  "global_step": 5000,
+  "source_checkpoint": "epoch_0-step_5000.ckpt",
+  "format": "safetensors",
+  "tensors": 320
+}

re10k-24view/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2dd5824a0465883f2e8eb69d32dd8eb763e5bccac103eb9901bf9fc411701a47
+size 473786456