|
--- |
|
license: apache-2.0 |
|
language: |
|
- en |
|
metrics: |
|
- rouge |
|
tags: |
|
- nanoT5 |
|
datasets: |
|
- allenai/c4 |
|
--- |
|
|
|
[Google's T5-v1.1-base](https://huggingface.co/google/t5-v1_1-base) pre-trained for 2^16 steps in [nanoT5](https://github.com/PiotrNawrot/nanoT5) library for efficient pre-training. |
|
|
|
For more details about the model refer to the original [paper](https://arxiv.org/pdf/2002.05202.pdf) and original [model weights](https://huggingface.co/google/t5-v1_1-base). |
|
|
|
This checkpoint was pre-trained on a single GPU for 20 hours. |
|
It can be further fine-tuned on SuperNatural-Instructions dataset to achieve 95% of the performance of the same model pre-trained through "a combination of model and data parallelism [...] on slices of Cloud TPU Pods", each with 1024 TPUs. |
