tensorflow / tensorflow/tflite-support
MobileNetV3 quantization
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 441
- Forks
- 146
- PR merge metrics
- No merged PRs in 30d
Description
Hi! I'm trying to quantize MobileNetV3 with tflite, but int8-model performs very poor. I think, it is because of linear quantization, which is too simple method not appropriate for any weights distributions. What else can I try? Are you going to support logarithmic scale for quantization in the future?
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
No files or tests are named. Start by reviewing the TFLite int8 quantization path used for MobileNetV3 and compare its linear scaling with the reported model behavior. Done would require a supported alternative such as logarithmic scaling, with evidence that MobileNetV3 int8 quality improves.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp, tensorflow
- Domain
- machine-learning, mobile-dev
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100