Skip to content

RTX 5090 benchmark on coco val2017; readme guidance update#23

Open
AlfredMoore wants to merge 1 commit into
dataplayer12:mainfrom
AlfredMoore:feat/am/benchmark-rtx5090
Open

RTX 5090 benchmark on coco val2017; readme guidance update#23
AlfredMoore wants to merge 1 commit into
dataplayer12:mainfrom
AlfredMoore:feat/am/benchmark-rtx5090

Conversation

@AlfredMoore

Copy link
Copy Markdown

Benchmark for RTX 5090 w Intel Ultra 7

Hardware HF+PyTorch TensorRT+CUDA Speedup Notes
RTX 5090 120.9 ms 24.8 ms 4.88x COCO val2017, 100% GPU UTL
image

Copilot AI review requested due to automatic review settings June 14, 2026 05:59

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Updates README documentation to reflect a new GPU benchmark entry and switches the documented ONNX export / TensorRT build instructions to a “dynamic” ONNX artifact.

Changes:

  • Added RTX 5090 latency numbers and environment note to the benchmark table.
  • Updated ONNX export output filename reference from sam3_static.onnx to sam3_dynamic.onnx.
  • Updated the trtexec example to build from the dynamic ONNX file.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment thread README.md
python python/onnxexport.py
```
This produces `onnx_weights/sam3_static.onnx` plus external weight shards.
This produces `onnx_weights/sam3_dynamic.onnx` plus external weight shards.
Comment thread README.md
5) Build a TensorRT engine
```bash
trtexec --onnx=onnx_weights/sam3_static.onnx --saveEngine=sam3_fp16.plan --fp16 --verbose
trtexec --onnx=onnx_weights/sam3_dynamic.onnx --saveEngine=sam3_fp16.plan --fp16 --verbose
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants