File size: 423 Bytes
7f25423
 
 
 
b806bfd
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
---
library_name: transformers
tags: []
---
Install: https://github.com/KONAKONA666/q8_kernels

# Usage

```python
import torch
from q8_kernels.models.T5EncoderFP8 import T5EncoderModelFP8
text_encoder = T5EncoderModelFP8.from_pretrained(
    "konakona/t5xxl_encoder_fp8", torch_dtype=torch.bfloat16
)
```
Needed dX for token training(PTI and textual inversion) in LTX. \
dX is in bf16, calculations in FP8 \
Needs Ada GPU