shtif
/

distilhubert-finetuned-gtzan

Audio Classification

Generated from Trainer

Inference Endpoints

Model card Files Files and versions Metrics Training metrics Community

shtif commited on Aug 8, 2023

Commit

e0841bb

•

1 Parent(s): 39abeea

update model card README.md

Files changed (1) hide show

README.md +8 -23

README.md CHANGED Viewed

@@ -18,8 +18,8 @@ should probably proofread and complete it, then remove this comment. -->
 This model is a fine-tuned version of [ntu-spml/distilhubert](https://huggingface.co/ntu-spml/distilhubert) on the GTZAN dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.9594
-- Accuracy: 0.83
 ## Model description
@@ -45,33 +45,18 @@ The following hyperparameters were used during training:
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 20
 - mixed_precision_training: Native AMP
 ### Training results
 | Training Loss | Epoch | Step | Validation Loss | Accuracy |
 |:-------------:|:-----:|:----:|:---------------:|:--------:|
-| 1.874         | 1.0   | 113  | 1.8949          | 0.42     |
-| 1.2872        | 2.0   | 226  | 1.3293          | 0.57     |
-| 0.9764        | 3.0   | 339  | 0.9030          | 0.72     |
-| 0.5805        | 4.0   | 452  | 0.6561          | 0.83     |
-| 0.4618        | 5.0   | 565  | 0.5127          | 0.87     |
-| 0.1487        | 6.0   | 678  | 0.7336          | 0.77     |
-| 0.1542        | 7.0   | 791  | 0.5496          | 0.84     |
-| 0.267         | 8.0   | 904  | 0.6534          | 0.85     |
-| 0.037         | 9.0   | 1017 | 0.7327          | 0.85     |
-| 0.0089        | 10.0  | 1130 | 1.1979          | 0.76     |
-| 0.0436        | 11.0  | 1243 | 1.0857          | 0.82     |
-| 0.003         | 12.0  | 1356 | 0.9266          | 0.84     |
-| 0.0019        | 13.0  | 1469 | 0.9791          | 0.84     |
-| 0.0017        | 14.0  | 1582 | 0.9259          | 0.84     |
-| 0.0015        | 15.0  | 1695 | 0.9836          | 0.83     |
-| 0.0014        | 16.0  | 1808 | 1.0018          | 0.83     |
-| 0.0013        | 17.0  | 1921 | 0.9896          | 0.83     |
-| 0.0012        | 18.0  | 2034 | 0.9836          | 0.84     |
-| 0.0012        | 19.0  | 2147 | 0.9759          | 0.84     |
-| 0.0011        | 20.0  | 2260 | 0.9594          | 0.83     |
 ### Framework versions

 This model is a fine-tuned version of [ntu-spml/distilhubert](https://huggingface.co/ntu-spml/distilhubert) on the GTZAN dataset.
 It achieves the following results on the evaluation set:
+- Loss: 0.6216
+- Accuracy: 0.85
 ## Model description
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_ratio: 0.1
+- num_epochs: 5
 - mixed_precision_training: Native AMP
 ### Training results
 | Training Loss | Epoch | Step | Validation Loss | Accuracy |
 |:-------------:|:-----:|:----:|:---------------:|:--------:|
+| 0.4112        | 1.0   | 113  | 0.6618          | 0.83     |
+| 0.1451        | 2.0   | 226  | 0.6944          | 0.82     |
+| 0.1972        | 3.0   | 339  | 0.5561          | 0.86     |
+| 0.0062        | 4.0   | 452  | 0.5888          | 0.88     |
+| 0.0058        | 5.0   | 565  | 0.6216          | 0.85     |
 ### Framework versions