Primus-Turbo

PyTorchConference2025_GithubRepos.json

@@ -985,6 +985,8 @@
   {
     "repo_name": "Primus-Turbo",
     "repo_link": "https://github.com/AMD-AGI/Primus-Turbo",
+    "category": "training framework",
+    "github_about_section": "Primus-Turbo is a high-performance acceleration library dedicated to large-scale model training on AMD GPUs. Built and optimized for the AMD ROCm platform, it covers the full training stack — including core compute operators (GEMM, Attention, GroupedGEMM), communication primitives, optimizer modules, low-precision computation (FP8), and compute–communication overlap kernels.",
     "contributors_all": 12,
     "contributors_2025": 12,
     "contributors_2024": 0,
@@ -1005,7 +1007,7 @@
   {
     "repo_name": "FTorch",
     "repo_link": "https://github.com/Cambridge-ICCS/FTorch",
-    "category": "wrapper",
+    "category": "middleware",
     "github_about_section": "A library for directly calling PyTorch ML models from Fortran.",
     "homepage_link": "https://cambridge-iccs.github.io/FTorch",
     "github_topic_closest_fit": "machine-learning",