Update README.md

README.md

@@ -42,7 +42,7 @@ We use the pretrained [`sentence-transformers/all-mpnet-base-v2`](https://huggin
  We use the initial model without modifying its architecture or pre-trained model parameters. 
  However, we reduce the processed sequence length to 256 to reduce the training time of the model. 
  Regarding other hyperparameters, we use the same interaction data batch size of 1024; we use the negative sampling parameter m = 10000.
- We use constant learning rate of 1e-5, and we train the model for ten epochs.
+ We use constant learning rate of 1e-5, and we train the model for five epochs.
  We finetuned our model on the MovieLens-20M dataset. For details please see our paper (link TBA).
  
  For item ids used during training please see (links TBA).