tensorflow / tensorflow/model-optimization
Input and resource quantization
Open
Nobody has claimed this yet.
feature request
- Dominant language
- Python
- Stars
- 1.6k
- Forks
- 349
- Avg merge
- 3d 2h
- Merged PRs (30d)
- 1
Description
System information
- TensorFlow version (you are using): tensorflow 2.9.1
- Are you willing to contribute it (Yes/No): No
Motivation
The input and resource need interface for customer layer quantization,
Normalization is a basic structure in the DNN, but we cannot quantize it easily.
- No input quantization interface in customer configuration, it will make quantization mul-add structure in float32 instead of int8 in tflite file.
- No resource quantization interface in customer configuration, it will make batchnorm - (running mean, running variance) wo/ QAT.
Contributor guide
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
The issue names no files, tests, or entry points. Start by locating customer quantization configuration and the TensorFlow Lite conversion and QAT paths; compare how input tensors and batch-normalization resources are currently handled. Done should include agreed interfaces and coverage showing that the relevant operations can be quantized as intended.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- keras, python, tensorflow
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100