Implementation of the PyTorch model compiler

[Sahil2004]

Description

We need to implement a compiler for a PyTorch model. This compiler should be capable of converting a PyTorch model into a format suitable for deployment in various environments. The primary objectives are to ensure compatibility, optimize performance, and facilitate integration.

Requirements

1. Model Parsing:

   • Develop a parser to read and interpret the PyTorch model architecture and parameters.
   • Ensure support for common layers and operations used in PyTorch.

2. Optimization:

   • Implement techniques to optimize the model for faster inference and reduced memory usage.
   • Include support for quantization, pruning, and other optimization strategies.

Tasks

1. Model Parsing

   • [ ] Develop a function to parse PyTorch model architecture.
   • [ ] Implement support for common PyTorch layers (Conv2d, Linear, etc.).
   • [ ] Handle model parameters and configurations.

2. Optimization

   • [ ] Implement model quantization.
   • [ ] Develop pruning strategies.
   • [ ] Explore other optimization techniques and integrate them.

Additional Information

• Target Environment: The compiler should support both CPU and GPU environments.
• Supported Models: Focus on commonly used models such as ResNet, MobileNet, etc.
• Performance Metrics: Measure and report metrics such as inference time, memory usage, and accuracy.

References

• PyTorch Documentation

---

Please comment if you have any suggestions or questions regarding this implementation plan.

[HarshitVashisht11]

Can i work on this 🙄

[Sahil2004]

Can i work on this 🙄

Alright. You are assigned this issue. Looking forward to a PR asap following all the guidelines :)

[HarshitVashisht11]

🫡

[HarshitVashisht11]

Like if we want to generate test cases for a compiler so why can't we use already build pytorch compilers instead of making a new compiler