Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- Phando/uspto-50k
|
| 4 |
+
metrics:
|
| 5 |
+
- accuracy
|
| 6 |
+
pipeline_tag: text-classification
|
| 7 |
+
tags:
|
| 8 |
+
- chemistry
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
This [ChemBERTa-v2](https://huggingface.co/seyonec/ChemBERTa_zinc250k_v2_40k) checkpoint was fine-tuned on the [USPTO-50k](https://huggingface.co/datasets/Phando/uspto-50k) dataset for sequence classification.
|
| 12 |
+
|
| 13 |
+
Specifically, the objective is to predict the reaction class label, and the input is either (canonicalized) all reactant SMILES or all product SMILES (separated by ".").
|
| 14 |
+
|
| 15 |
+
- Train/Test split: 0.99/0.01
|
| 16 |
+
|
| 17 |
+
- Evaluation results:
|
| 18 |
+
- Accuracy: 87.11%
|
| 19 |
+
- Loss: 0.4272
|
| 20 |
+
|
| 21 |
+
- Fine-tuning hyperparameters:
|
| 22 |
+
- seed = 233
|
| 23 |
+
- batch-size = 128
|
| 24 |
+
- num_epochs = 5 (but early stopped at epoch 4)
|
| 25 |
+
- learning_rate = 5e-4
|
| 26 |
+
- warmup_steps = 64
|
| 27 |
+
- weight_decay = 0.01
|