Prebuilt TensorRT 10.7 engine for Jetson Orin (sm_87)

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+da2_metric_hypersim_vits_364x448_trt100700_sm87.engine filter=lfs diff=lfs merge=lfs -text

README.md

@@ -15,5 +15,8 @@ Depth Anything V2 Small, metric, Hypersim (indoor, max depth 20 m), converted fr
 - `dinov2_vits14.safetensors`, `da2_head_vits.safetensors`: the encoder and the DPT head, for candle
   (converted by `tools/convert_weights.py`; includes the unused ImageNet classifier head candle's dinov2 loads).
 - `da2_metric_hypersim_vits_364x448.onnx`: an ONNX export with a fixed 364x448 input, for TensorRT on Jetsons.
+- `da2_metric_hypersim_vits_364x448_trt100700_sm87.engine`: that ONNX as an fp16 TensorRT 10.7 engine, built on a
+  Jetson AGX Orin (sm_87, JetPack 6 / L4T R36.3), so Orins skip the minutes-long engine build. Other TensorRT
+  versions or GPUs reject it and the crate builds its own from the ONNX.
 
 The crate pins these files by sha256 in its `model.json`. Weights are Apache-2.0, as the upstream Small model.

da2_metric_hypersim_vits_364x448_trt100700_sm87.engine

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64fab3e772a7d7cb8d5728d84b2e497db157065b97db5a33e72701a07e4ef6fe
+size 53823004