Update README.md
Browse files
README.md
CHANGED
@@ -83,6 +83,13 @@ https://github.com/nicolay-r/Reasoning-for-Sentiment-Analysis-Framework
|
|
83 |
Google-colab notebook for reproduction:
|
84 |
https://colab.research.google.com/github/nicolay-r/Reasoning-for-Sentiment-Analysis-Framework/blob/main/Reasoning_for_Sentiment_Analysis_Framework.ipynb
|
85 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
The overall training process took **3 epochs**.
|
87 |
|
88 |
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64e62d11d27a8292c3637f86/JwCP0EIe6q1VVdNrTzPQl.png)
|
|
|
83 |
Google-colab notebook for reproduction:
|
84 |
https://colab.research.google.com/github/nicolay-r/Reasoning-for-Sentiment-Analysis-Framework/blob/main/Reasoning_for_Sentiment_Analysis_Framework.ipynb
|
85 |
|
86 |
+
**Setup:** `Flan-T5-xl` up to **64 tokens**, 4-batch size.
|
87 |
+
|
88 |
+
**GPU:** NVidia-A100, **bfloat16**, ~ 30 min/epoch
|
89 |
+
|
90 |
+
> NOTE: could not be fitted in 40GB VRAM. FP16 mode lead to `Nan` loss
|
91 |
+
|
92 |
+
|
93 |
The overall training process took **3 epochs**.
|
94 |
|
95 |
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64e62d11d27a8292c3637f86/JwCP0EIe6q1VVdNrTzPQl.png)
|