View the DistilBERT benchmark results

Open the benchmark dashboard

In addition to the GPT-2 results, the results directory should now contain three DistilBERT files:

• distilbert_pytorch.json
• distilbert_portable.json
• distilbert_xnnpack.json

The same dashboard can display the GPT-2 and DistilBERT results together. If it is still running from the GPT-2 section, refresh the browser. Otherwise, restart it from the Raspberry Pi:

    

        
        
course_support/serve_results.py \
  --results ~/executorch_on_arm/module-1-transformer-inference-with-xnnpack/results \
  --host 127.0.0.1 \
  --port 8080

    

Open the same local URL on your development machine:

    

        
        
http://localhost:8080

    

Example DistilBERT results are provided below, and yours will likely show a similar pattern, though are unlikely to be identical:

In the example data, you will see the same pattern repeated from the GPT-2 tests.

XNNPACK reduced mean latency from 190.75 ms for PyTorch eager execution, to 80.94 ms, a speedup of approximately 2.4×. The single run of the portable kernel artifact took around 20.7 seconds.