Philip May
commited on
Commit
·
4519bfb
1
Parent(s):
30e5e4f
Update README.md
Browse files
README.md
CHANGED
|
@@ -12,7 +12,12 @@ tags:
|
|
| 12 |
- Ph. S. (company)
|
| 13 |
|
| 14 |
# Evaluation
|
| 15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
|
| 17 |
# Tips for training on GPUs
|
| 18 |
This model is too big to fit on a normal 16GB GPU in FP32 mode.
|
|
|
|
| 12 |
- Ph. S. (company)
|
| 13 |
|
| 14 |
# Evaluation
|
| 15 |
+
Evaluation was done on a summarization task with:
|
| 16 |
+
- train data: [Swisstext](https://www.swisstext.org/2019/shared-task/german-text-summarization-challenge.html)
|
| 17 |
+
- test data: [MLSUM](https://huggingface.co/datasets/mlsum)
|
| 18 |
+
- GPUs: 4 (V100)
|
| 19 |
+
|
| 20 |
+
For details see <https://github.com/GermanT5/german-t5-eval>
|
| 21 |
|
| 22 |
# Tips for training on GPUs
|
| 23 |
This model is too big to fit on a normal 16GB GPU in FP32 mode.
|