addressing numpy2 compatibility issue (#2995)

Author: ZailiWang

docs/tutorials/known_issues.md

@@ -42,6 +42,16 @@ Troubleshooting
   - **Cause**: PyTorch FX limitation.
   - **Solution**: You can avoid this error by calling `m = ipex.optimize(m, level="O0")`, which doesn't apply ipex optimization, or disable `conv+bn` folding by calling `m = ipex.optimize(m, level="O1", conv_bn_folding=False)`.
 
+## Package Compatibility
+
+- **Problem**: Error raised during the import of `numpy`.
+  - **Cause**: `numpy` has some API update in latest 2.0.0 release, but it comes incompatible with some other packages.
+  - **Solution**: Reinstall a lagacy version of `numpy` with the command:
+
+  ```
+  python -m pip install numpy==1.26.4 --force-reinstall
+  ```
+
 ## Performance Regression
 
 - Some models may experience performance regression comparing to 2.0.x due to deprecation of the NNC feature in PyTorch\*.

examples/cpu/inference/python/llm/tools/env_setup.sh

@@ -226,3 +226,6 @@ if [ $((${MODE} & 0x01)) -ne 0 ]; then
     fi
     ln -s ../prompt.json
 fi
+
+# Install a legacy numpy version to avoid compatibility issue with deepspeed
+python -m pip install numpy==1.26.4 --force-reinstall