Optimize exposure autoencoder distributed training

README.md

@@ -73,3 +73,16 @@ compatible with the default gated `TimesNetExposureEncoder`, and the channel
 normalization statistics needed when the encoder is attached to DeepHealth.
 Multi-GPU pretraining follows the main trainer interface: add
 `--data_parallel --gpu_ids 0,1,2,3`.
+For efficient multi-GPU training, launch one process per GPU with DDP:
+
+```bash
+torchrun --standalone --nproc_per_node=4 train_exposure_autoencoder.py \
+  --exposure_cache_dir ukb_exposure_cache \
+  --batch_size 128
+```
+
+In DDP mode, `--batch_size` is the global batch size and must be divisible by
+the number of processes.
+Training-channel statistics are cached at
+`<exposure_cache_dir>/train_channel_stats.npz`; use
+`--recompute_channel_stats` only when a forced refresh is needed.