Merge branch 'main' into dev/yk_perf_model

Author: yuankaichen-amd

README.md

@@ -5,6 +5,7 @@
 ---
 
 ## 🆕 What's New
+- **[2025/08/22]** Primus introduction [blog](https://rocm.blogs.amd.com/software-tools-optimization/primus/README.html).
 - **[2025/06/18]** Added TorchTitan backend support.
 - **[2025/05/16]** Added benchmark suite for performance evaluation across models and hardware.
 - **[2025/04/18]** Added [Preflight](./tools/preflight/README.md) cluster sanity checker to verify environment readiness.
@@ -20,7 +21,7 @@
 |--------------|------|--------------|-----------------------------|
 | [**Primus-LM**](https://github.com/AMD-AGI/Primus)         | End-to-end training framework | - Supports multiple training backends (Megatron, TorchTitan, etc.)<br>- Provides high-performance, scalable distributed training<br>- Deeply integrates with Turbo and Safe | - Can invoke Primus-Turbo kernels and modules<br>- Runs on top of Primus-Safe for stable scheduling |
 | [**Primus-Turbo**](https://github.com/AMD-AGI/Primus-Turbo)         | High-performance operators & modules | - Provides common LLM training operators (FlashAttention, GEMM, Collectives, GroupedGemm, etc.)<br>- Modular design, directly pluggable into Primus-LM<br>- Optimized for different architectures and precisions | - Built on [**AITER**](https://github.com/ROCm/aiter), [**CK**](https://github.com/ROCm/composable_kernel), [**hipBLASLt**](https://github.com/ROCm/hipBLASLt), [**Triton**](https://github.com/ROCm/triton)  and other operator libraries<br>- Can be enabled via configuration inside Primus-LM |
-| **Primus-SaFE** (Coming soon)         | Stability & platform layer | - Cluster sanity check and benchmarking<br>- Kubernets scheduling with topology awareness<br>- Fault tolerance<br>- Stability enhancements | - Building a training platform based on the K8s and Slurm ecosystem |
+| [**Primus-SaFE**](https://github.com/AMD-AGI/Primus-SaFE)         | Stability & platform layer | - Cluster sanity check and benchmarking<br>- Kubernets scheduling with topology awareness<br>- Fault tolerance<br>- Stability enhancements | - Building a training platform based on the K8s and Slurm ecosystem |
 
 ---