Update README.md

README.md

@@ -87,38 +87,7 @@ Optimising the threshold per label to optimise the F1 metric, the metrics (evalu
 | surprise       | 0.301 |     0.235 |  0.418 |     141 |      0.10 |
 | neutral        | 0.626 |     0.519 |  0.786 |    1787 |      0.30 |
 
-Using a fixed threshold of 0.5 to convert the scores to binary predictions for each label, the metrics (evaluated on the go_emotions test split) are:
-
-|               |    f1 | precision | recall | support | threshold |
-| ------------- | ----- | --------- | ------ | ------- | --------- |
-|admiration     | 0.443 |     0.722 |  0.319 |     504 |       0.5 |
-|amusement      | 0.364 |     0.805 |  0.235 |     264 |       0.5 |
-|anger          | 0.100 |     0.478 |  0.056 |     198 |       0.5 |
-|annoyance      | 0.012 |     0.667 |  0.006 |     320 |       0.5 |
-|approval       | 0.082 |     0.882 |  0.043 |     351 |       0.5 |
-|caring         | 0.118 |     0.500 |  0.067 |     135 |       0.5 |
-|confusion      | 0.107 |     0.600 |  0.059 |     153 |       0.5 |
-|curiosity      | 0.242 |     0.550 |  0.155 |     284 |       0.5 |
-|desire         | 0.204 |     0.667 |  0.120 |      83 |       0.5 |
-|disappointment | 0.026 |     1.000 |  0.013 |     151 |       0.5 |
-|disapproval    | 0.084 |     0.600 |  0.045 |     267 |       0.5 |
-|disgust        | 0.243 |     0.720 |  0.146 |     123 |       0.5 |
-|embarrassment  | 0.217 |     0.556 |  0.135 |      37 |       0.5 |
-|excitement     | 0.037 |     0.333 |  0.019 |     103 |       0.5 |
-|fear           | 0.466 |     0.711 |  0.346 |      78 |       0.5 |
-|gratitude      | 0.757 |     0.915 |  0.645 |     352 |       0.5 |
-|grief          | 0.286 |     0.200 |  0.500 |       6 |       0.5 |
-|joy            | 0.197 |     0.818 |  0.112 |     161 |       0.5 |
-|love           | 0.519 |     0.805 |  0.382 |     238 |       0.5 |
-|nervousness    | 0.293 |     0.333 |  0.261 |      23 |       0.5 |
-|optimism       | 0.260 |     0.784 |  0.156 |     186 |       0.5 |
-|pride          | 0.444 |     0.545 |  0.375 |      16 |       0.5 |
-|realization    | 0.014 |     0.500 |  0.007 |     145 |       0.5 |
-|relief         | 0.154 |     0.500 |  0.091 |      11 |       0.5 |
-|remorse        | 0.449 |     0.606 |  0.357 |      56 |       0.5 |
-|sadness        | 0.297 |     0.744 |  0.186 |     156 |       0.5 |
-|surprise       | 0.042 |     1.000 |  0.021 |     141 |       0.5 |
-|neutral        | 0.528 |     0.649 |  0.445 |    1787 |       0.5 |
+The thesholds are stored in `thresholds.json`.
 
 ### Use with ONNXRuntime
 
@@ -145,6 +114,8 @@ print(outputs[0])
 > surprise
 print(preds_onnx[0])
 > array([[0.97136074, 0.02863926]], dtype=float32)
+
+# load thresholds.json and use that (per label) to convert the positive case score to a binary prediction
 ```
 
 ### Commentary on the dataset