accelerate

58c3605f - docs: update low-precision training docs to reflect MS-AMP deprecation (#3929)

docs: update low-precision training docs to reflect MS-AMP deprecation (#3929) - Add prominent deprecation warnings for MS-AMP in both usage and concept guides - Note specific compatibility issues (CUDA 12.x+, modern NCCL, PyTorch 2.2+) - Recommend TransformersEngine and torchao as actively maintained alternatives - Update code examples to prefer TE/torchao over MS-AMP - Reorder Further Reading links to de-emphasize MS-AMP Closes #3639