Update README.md
Browse files
README.md
CHANGED
@@ -16,7 +16,38 @@ base_model: westlake-repl/SaProt_35M_AF2
|
|
16 |
<!-- Provide a longer summary of what this model is. -->
|
17 |
This model is used to predict TM score give two SA protein pairs.
|
18 |
|
|
|
|
|
19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
|
22 |
|
|
|
16 |
<!-- Provide a longer summary of what this model is. -->
|
17 |
This model is used to predict TM score give two SA protein pairs.
|
18 |
|
19 |
+
### Model input type
|
20 |
+
SA sequence pair
|
21 |
|
22 |
+
### Performance
|
23 |
+
test_loss: 0.0031981547363102436
|
24 |
+
test_spearman: 0.8464066386222839
|
25 |
+
test_R2: 0.8718534111976624
|
26 |
+
test_pearson: 0.9338204860687256
|
27 |
+
|
28 |
+
### LoRA config
|
29 |
+
lora_dropout: 0.0
|
30 |
+
|
31 |
+
lora_alpha: 16
|
32 |
+
|
33 |
+
target_modules: ["query", "key", "value", "intermediate.dense", "output.dense"]
|
34 |
+
|
35 |
+
modules_to_save: ["classifier"]
|
36 |
+
|
37 |
+
### Training config
|
38 |
+
class: AdamW
|
39 |
+
|
40 |
+
betas: (0.9, 0.98)
|
41 |
+
|
42 |
+
weight_decay: 0.01
|
43 |
+
|
44 |
+
learning rate: 1e-4
|
45 |
+
|
46 |
+
epoch: 10
|
47 |
+
|
48 |
+
batch size: 4
|
49 |
+
|
50 |
+
precision: 16-mixed
|
51 |
|
52 |
|
53 |
|