2 -> 3

Looks like the 3-epoch model name ended with -2, but was changed to -3 (makes sense).
This makes the appropriate update.

README.md

@@ -9,7 +9,7 @@ language:
 # gpt4all-lora
 
 An autoregressive transformer trained on [data](https://huggingface.co/datasets/nomic-ai/gpt4all_prompt_generations) curated using [Atlas](https://atlas.nomic.ai/).
-This model is trained with four full epochs of training, while the related [gpt4all-lora-epoch-2 model](https://huggingface.co/nomic-ai/gpt4all-lora-epoch-2) is trained with three.
+This model is trained with four full epochs of training, while the related [gpt4all-lora-epoch-3 model](https://huggingface.co/nomic-ai/gpt4all-lora-epoch-3) is trained with three.
 Replication instructions and data: [https://github.com/nomic-ai/gpt4all](https://github.com/nomic-ai/gpt4all)
 
 ## Model Details