Deprecate compression examples

compression/README.md

@@ -1,12 +1,9 @@
-# DeepSpeed Model Compression examples
+# Compression examples
 
-Examples in this folder are helpful to try out some features and models that take advantage of the DeepSpeed compression library.
-
-A detailed tutorial for understanding and using DeepSpeed model compression features can be seen from here: https://www.deepspeed.ai/tutorials/model-compression/
+The legacy BERT, GPT-2, and CIFAR examples were removed because they depended
+on `deepspeed.compression`, which was removed in
+[deepspeedai/DeepSpeed#8490](https://github.com/deepspeedai/DeepSpeed/pull/8490).
 
 | Example | Description |
 | --- | --- |
-| [bert](bert) | Quantization, pruning and layer reduction on BERT (ZeroQuant, XTC) |
-| [gpt2](gpt2) | ZeroQuant post-training quantization on GPT-2 |
-| [cifar](cifar) | Channel pruning and quantization on a CIFAR ResNet |
 | [reasoning_aware_compression](reasoning_aware_compression) | One-shot pruning of reasoning LLMs (DeepSeek-R1 distills, Qwen3) calibrated on their own chain-of-thought traces — [RAC, ICLR 2026](https://arxiv.org/abs/2509.12464) |