README.md

## Variants

* Default (`fp32`)
* `fp16`
* `bf16`
* `fp8_e4m3`

## Usage

```python
import torch
from transformers import T5EncoderModel

model = T5EncoderModel.from_pretrained("hlky/t5-v1_1-xxl-encoder", variant="fp16", torch_dtype=torch.float16)
```