T5 Efficient Large Dm2000

🚀 T5-Efficient-LARGE-DM2000 (Deep-Narrow version)

T5-Efficient-LARGE-DM2000 is a variation of Google's original T5, following the T5 model architecture. It is a pretrained-only checkpoint, released with the paper Scale Efficiently: Insights from Pre-training and Fine-tuning Transformers by Yi Tay, Mostafa Dehghani, Jinfeng Rao, William Fedus, Samira Abnar, Hyung Won Chung, Sharan Narang, Dani Yogatama, Ashish Vaswani, Donald Metzler. This model demonstrates that a Deep-Narrow architecture can be more favorable for downstream performance compared to other models with a similar parameter count.

🚀 Quick Start

This model is a pretrained-only checkpoint and needs to be fine-tuned for practical use. It was pretrained in English, so it's suitable for English NLP tasks. You can refer to the following examples for fine-tuning:

PyTorch

• Summarization
• Question Answering
• Text Classification - Note: You need to slightly adapt the training example to work with an encoder-decoder model.

Tensorflow

• Summarization
• Text Classification - Note: You need to slightly adapt the training example to work with an encoder-decoder model.

JAX/Flax

• Summarization
• Text Classification - Note: You need to slightly adapt the training example to work with an encoder-decoder model.

✨ Features

The paper indicates that a Deep-Narrow model architecture is more favorable for downstream performance compared to other model architectures of similar parameter count. To quote the paper:

We generally recommend a DeepNarrow strategy where the model’s depth is preferentially increased before considering any other forms of uniform scaling across other dimensions. This is largely due to how much depth influences the Pareto-frontier as shown in earlier sections of the paper. Specifically, a tall small (deep and narrow) model is generally more efficient compared to the base model. Likewise, a tall base model might also generally more efficient compared to a large model. We generally find that, regardless of size, even if absolute performance might increase as we continue to stack layers, the relative gain of Pareto-efficiency diminishes as we increase the layers, converging at 32 to 36 layers. Finally, we note that our notion of efficiency here relates to any one compute dimension, i.e., params, FLOPs or throughput (speed). We report all three key efficiency metrics (number of params, FLOPS and speed) and leave this decision to the practitioner to decide which compute dimension to consider.

📚 Documentation

🔧 Technical Details

Model Architecture

This model checkpoint - t5-efficient-large-dm2000 - is of model type Large with the following variations:

• dm is 2000

It has 1475.39 million parameters and thus requires ca. 5901.57 MB of memory in full precision (fp32) or 2950.78 MB of memory in half precision (fp16 or bf16).

A summary of the original T5 model architectures can be seen here:

The following abbreviations are used:

If a model checkpoint has no specific el or dl, then both the number of encoder- and decoder layers correspond to nl.

Pre - Training

The checkpoint was pretrained on the Colossal, Cleaned version of Common Crawl (C4) for 524288 steps using the span-based masked language modeling (MLM) objective.

More Information

We strongly recommend the reader to go carefully through the original paper Scale Efficiently: Insights from Pre-training and Fine-tuning Transformers to get a more nuanced understanding of this model checkpoint. As explained in the following issue, checkpoints including the sh or skv model architecture variations have not been ported to Transformers as they are probably of limited practical usage and are lacking a more detailed description. Those checkpoints are kept here as they might be ported potentially in the future.

📄 License

This model is released under the apache-2.0 license.