Create README.md (#1)
Browse files- Create README.md (ed994783d01629e1c3b167331a8093b1e759ba0b)
- fixed typos, updated model size values (ac04cb38d8d6e6aad97dec51a3a6c6d09c83daaa)
README.md
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- lambada
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
library_name: transformers
|
| 8 |
+
pipeline_tag: text-generation
|
| 9 |
+
tags:
|
| 10 |
+
- text-generation-inference
|
| 11 |
+
- causal-lm
|
| 12 |
+
- int8
|
| 13 |
+
- tensorrt
|
| 14 |
+
- ENOT-AutoDL
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# INT8 GPT-J 6B
|
| 18 |
+
|
| 19 |
+
GPT-J 6B is a transformer model trained using Ben Wang's [Mesh Transformer JAX](https://github.com/kingoflolz/mesh-transformer-jax/). "GPT-J" refers to the class of model, while "6B" represents the number of trainable parameters.
|
| 20 |
+
|
| 21 |
+
This repository contains TensorRT engines with mixed precission int8 + fp32. You can find prebuilt engines for next GPUs:
|
| 22 |
+
* RTX 4090
|
| 23 |
+
* RTX 3080 Ti
|
| 24 |
+
* RTX 2080 Ti
|
| 25 |
+
|
| 26 |
+
ONNX model generated by [ENOT-AutoDL](https://pypi.org/project/enot-autodl/) and will be published soon.
|
| 27 |
+
|
| 28 |
+
## Test result
|
| 29 |
+
|
| 30 |
+
| |INT8|FP32|
|
| 31 |
+
|---|:---:|:---:|
|
| 32 |
+
| **Lambada Acc** |78.50%|79.54%|
|
| 33 |
+
| **Model size (GB)** |8.5|24.2|
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
## How to use
|
| 37 |
+
|
| 38 |
+
Example of inference and accuracy test published on github:
|
| 39 |
+
```shell
|
| 40 |
+
git clone https://github.com/ENOT-AutoDL/demo-gpt-j-6B-tensorrt-int8
|
| 41 |
+
```
|