AMD FP8 Training Optimizations Merged Into PyTorch TorchTitan and TorchAO

AMD and Meta/PyTorch engineers have upstreamed FP8 training optimizations into pytorch/AO and pytorch/TorchTitan, enabling AMD Instinct GPU support with competitive FP8 performance out of the box¹. On Llama3-8B, FP8 training delivers a 13.4% throughput gain over BF16. For DeepSeek-V3 671B MoE shapes, fused Triton quantization kernels recovered 89% of FP8 quantization overhead, with individual kernel optimizations delivering up to a 6.2× speedup. At the PyTorch Conference 2025, the team demonstrated linear scaling beyond 1,000 GPUs on AMD Instinct clusters.