Quick and easy profiling of CoreML models in Python!
#convert model using guide: https://apple.github.io/coremltools/docs-guides/source/introductory-quickstart.html
#or download one from the 🤗 Hub
from coremlprofiler import CoreMLProfiler
from huggingface_hub import snapshot_download
repo_path = snapshot_download(repo_id="FL33TW00D-HF/test-st", local_dir="./")
model_path = repo_path + "/sentence_transformer_all-MiniLM-L6-v2.mlpackage"
# Create the Profiler
profiler = CoreMLProfiler(model_path)
# Print your device usage
print(profiler.device_usage_summary_chart())
# Print the operator compatibility report
print(profiler.operator_compatibility_report())
For more detailed reasons about why an operation could not be run on a compute unit, try out decoreml
Profile a local model:
coremlprofile path/to/my/model.mlpackageDownload from Hugging Face and profile:
coremlprofile --hf_repo="FL33TW00D-HF/test-st" sentence_transformer_all-MiniLM-L6-v2.mlpackage