docs: fix outdated link to TF32 explanation (#32947)

fix outdated link

src/transformers/training_args.py

@@ -416,7 +416,7 @@ class TrainingArguments:
         tf32 (`bool`, *optional*):
             Whether to enable the TF32 mode, available in Ampere and newer GPU architectures. The default value depends
             on PyTorch's version default of `torch.backends.cuda.matmul.allow_tf32`. For more details please refer to
-            the [TF32](https://huggingface.co/docs/transformers/performance#tf32) documentation. This is an
+            the [TF32](https://huggingface.co/docs/transformers/perf_train_gpu_one#tf32) documentation. This is an
             experimental API and it may change.
         local_rank (`int`, *optional*, defaults to -1):
             Rank of the process during distributed training.