Update README.md

README.md

@@ -10,6 +10,7 @@ datasets:
 
 ---
 
+- - [Version ✨v1✨](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1): August 25th, 2022 (*[full](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1) and [half-precision weights](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1-half)*, at step 1M)
 - [Version v1beta3](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1beta3): July 22nd, 2022 (*[full](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1beta3) and [half-precision weights](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1beta3-half)*, at step 850k)
 - [Version v1beta2](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1beta2): June 6th, 2022 (*[full](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1beta2) and [half-precision weights](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1beta2-half)*, at step 616k)
 - [Version v1beta1](https://huggingface.co/bertin-project/bertin-gpt-j-6B/tree/v1beta1-half): April 28th, 2022 (*half-precision weights only*, at step 408k)
@@ -63,7 +64,7 @@ BERTIN-GPT-J-6B was finetuned on [mC4-es-sampled (gaussian)](https://huggingface
 
 ## Training procedure
 
-This model was finetuned for 40 billion tokens (40,384,790,528) over 616,000 steps on a single TPU v3-8 VM. It was trained as an autoregressive language model, using cross-entropy loss to maximize the likelihood of predicting the next token correctly.
+This model was finetuned for ~65 billion tokens (65,536,000,000) over 1,000,000 steps on a single TPU v3-8 VM. It was trained as an autoregressive language model, using cross-entropy loss to maximize the likelihood of predicting the next token correctly. Training took roughly 6 months.
 
 ## Intended Use and Limitations