Update README.md
Browse files
README.md
CHANGED
@@ -26,11 +26,11 @@ Google's mt5-base fine-tuned in Japanese to summarize patent claims in a limited
|
|
26 |
- "-small" has been trained on 20,000 text pairs only. (Training data seems to be too little.)
|
27 |
- dataset: *
|
28 |
- prefix: "patent claim summarization: " (notice: single task trained.)
|
29 |
-
- NOTICE: preferable to set "no_repeat_ngram_size" on model.generate().
|
30 |
|
31 |
#参考
|
32 |
|
33 |
- https://huggingface.co/blog/how-to-generate
|
34 |
-
-
|
|
|
35 |
|
36 |
**check in progress**
|
|
|
26 |
- "-small" has been trained on 20,000 text pairs only. (Training data seems to be too little.)
|
27 |
- dataset: *
|
28 |
- prefix: "patent claim summarization: " (notice: single task trained.)
|
|
|
29 |
|
30 |
#参考
|
31 |
|
32 |
- https://huggingface.co/blog/how-to-generate
|
33 |
+
- 前処理が最適ではなかった。修正する。
|
34 |
+
- 次回は、テーマに沿った要約の改変をおこなう。
|
35 |
|
36 |
**check in progress**
|