Training in progress, step 166

README.md

@@ -31,25 +31,25 @@ model-index:
       type: fineweb_c_eval
     metrics:
     - type: accuracy
-      value: 0.8335597826086957
+      value: 0.8376358695652174
       name: Accuracy
     - type: accuracy_threshold
-      value: 0.9343951344490051
+      value: 0.6088266372680664
       name: Accuracy Threshold
     - type: f1
-      value: 0.33268101761252444
+      value: 0.410958904109589
       name: F1
     - type: f1_threshold
-      value: 0.6988433599472046
+      value: 0.3380771279335022
       name: F1 Threshold
     - type: precision
-      value: 0.21935483870967742
+      value: 0.3105590062111801
       name: Precision
     - type: recall
-      value: 0.6882591093117408
+      value: 0.6072874493927125
       name: Recall
     - type: average_precision
-      value: 0.2441194063501111
+      value: 0.3509417036235269
       name: Average Precision
 ---
 
@@ -110,7 +110,7 @@ pairs = [
 ]
 scores = model.predict(pairs)
 print(scores)
-# [0.7461 0.8963 0.5265 0.5997 0.5913]
+# [0.4121 0.3208 0.0526 0.1956 0.3355]
 
 # Or rank different texts based on similarity to a single text
 ranks = model.rank(
@@ -161,13 +161,13 @@ You can finetune this model on your own dataset.
 
 | Metric                | Value      |
 |:----------------------|:-----------|
-| accuracy              | 0.8336     |
-| accuracy_threshold    | 0.9344     |
-| f1                    | 0.3327     |
-| f1_threshold          | 0.6988     |
-| precision             | 0.2194     |
-| recall                | 0.6883     |
-| **average_precision** | **0.2441** |
+| accuracy              | 0.8376     |
+| accuracy_threshold    | 0.6088     |
+| f1                    | 0.411      |
+| f1_threshold          | 0.3381     |
+| precision             | 0.3106     |
+| recall                | 0.6073     |
+| **average_precision** | **0.3509** |
 
 <!--
 ## Bias, Risks and Limitations
@@ -370,12 +370,22 @@ You can finetune this model on your own dataset.
 | 0.1735 | 72   | 1.2607        | -               | -                                |
 | 0.1952 | 81   | 1.2275        | -               | -                                |
 | 0.2    | 83   | -             | 1.3164          | 0.2441                           |
+| 0.2169 | 90   | 1.1550        | -               | -                                |
+| 0.2386 | 99   | 1.1969        | -               | -                                |
+| 0.2602 | 108  | 1.2051        | -               | -                                |
+| 0.2819 | 117  | 1.1031        | -               | -                                |
+| 0.3036 | 126  | 1.1517        | -               | -                                |
+| 0.3253 | 135  | 1.1449        | -               | -                                |
+| 0.3470 | 144  | 1.0654        | -               | -                                |
+| 0.3687 | 153  | 1.0757        | -               | -                                |
+| 0.3904 | 162  | 0.9787        | -               | -                                |
+| 0.4    | 166  | -             | 1.1451          | 0.3509                           |
 
 
 ### Training Time
-- **Training**: 1.2 minutes
-- **Evaluation**: 36.4 seconds
-- **Total**: 1.8 minutes
+- **Training**: 2.4 minutes
+- **Evaluation**: 1.2 minutes
+- **Total**: 3.6 minutes
 
 ### Framework Versions
 - Python: 3.12.12

eval/CrossEncoderClassificationEvaluator_fineweb_c_eval_results.csv

@@ -1,2 +1,3 @@
 epoch,steps,Accuracy,Accuracy_Threshold,F1,F1_Threshold,Precision,Recall,Average_Precision
 0.2,83,0.8335597826086957,0.93439513,0.33268101761252444,0.69884336,0.21935483870967742,0.6882591093117408,0.2441194063501111
+0.4,166,0.8376358695652174,0.60882664,0.410958904109589,0.33807713,0.3105590062111801,0.6072874493927125,0.3509417036235269

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e530136a08e1826eaf0bf913bf83a7e998dd2df53d492f794a427728a04b6902
+oid sha256:5198e0a994ddabb58b74e9c026fa1620ad8c31639d4cba5055d9750bca27116a
 size 1230138348