Summary of Changes

Hello @JyotinderSingh, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly enhances the Keras documentation by introducing a comprehensive guide on Activation-aware Weight Quantization (AWQ). The new guide provides a clear explanation of AWQ's mechanisms, practical implementation steps for Keras models, and detailed performance benchmarks. Furthermore, it integrates a comparative analysis with GPTQ, offering users a clear understanding of when to choose each quantization method based on their specific needs.

Highlights

• New AWQ Quantization Guide: Introduces a new guide (awq_quantization_in_keras.py, .ipynb, .md) detailing Activation-aware Weight Quantization (AWQ) for Keras and KerasHub models.
• AWQ Explanation and Usage: The guide explains the principles of AWQ, its workflow, and provides a practical example of quantizing a Gemma3CausalLM model using keras.quantizers.AWQConfig.
• Performance Benchmarking: Includes micro-benchmarks demonstrating the impact of AWQ on disk size, GPU memory, perplexity, and throughput for various models.
• AWQ vs. GPTQ Comparison: Adds a dedicated section comparing AWQ and GPTQ quantization methods, outlining their algorithms, speed, supported bit-widths, accuracy, memory usage, and calibration sensitivity, to both the new AWQ guide and the existing GPTQ guide.
• Guide Registration: The new AWQ guide has been registered in scripts/guides_master.py to ensure it appears in the documentation.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.