README.md and .gitattributes after training
Browse files
README.md
CHANGED
@@ -21,61 +21,19 @@ It achieves the following results on the evaluation set:
|
|
21 |
- Loss: 0.1385
|
22 |
- Accuracy: 0.9962
|
23 |
|
24 |
-
|
25 |
-
with the labels *down* and *on*.
|
26 |
-
Superb ks is in turn derived from (Speech Commands dataset v1.0)[https://www.tensorflow.org/datasets/catalog/speech_commands].
|
27 |
-
Train/validation/test splits are as in superb ks.
|
28 |
|
29 |
-
|
30 |
|
31 |
-
|
32 |
-
|
33 |
-
|
|
|
|
|
|
|
|
|
34 |
|
35 |
## Training procedure
|
36 |
-
Training used 'sbatch' on a cluster and the program [run_audio_classification.py](https://github.com/huggingface/transformers).
|
37 |
-
'down_on.sub' is below, start it with 'sbatch down_on.sub'.
|
38 |
-
|
39 |
-
'''
|
40 |
-
#!/bin/bash
|
41 |
-
#SBATCH -J down_on # Job name
|
42 |
-
#SBATCH -o down_on_%j.out # Name of stdout output log file (%j expands to jobID)
|
43 |
-
#SBATCH -e down_on_%j.err # Name of stderr output log file (%j expands to jobID)
|
44 |
-
#SBATCH -N 1 # Total number of nodes requested
|
45 |
-
#SBATCH -n 1 # Total number of cores requested
|
46 |
-
#SBATCH --mem=5000 # Total amount of (real) memory requested (per node)
|
47 |
-
#SBATCH -t 10:00:00 # Time limit (hh:mm:ss)
|
48 |
-
#SBATCH --partition=gpu # Request partition for resource allocation
|
49 |
-
#SBATCH --gres=gpu:1 # Specify a list of generic consumable resources (per node)
|
50 |
-
|
51 |
-
cd ~/ac_h
|
52 |
-
/home/mr249/env/hugh/bin/python run_audio_classification.py \
|
53 |
-
--model_name_or_path facebook/wav2vec2-base \
|
54 |
-
--dataset_name MatsRooth/down_on \
|
55 |
-
--output_dir wav2vec2-base_down_on \
|
56 |
-
--overwrite_output_dir \
|
57 |
-
--remove_unused_columns False \
|
58 |
-
--do_train \
|
59 |
-
--do_eval \
|
60 |
-
--fp16 \
|
61 |
-
--learning_rate 3e-5 \
|
62 |
-
--max_length_seconds 1 \
|
63 |
-
--attention_mask False \
|
64 |
-
--warmup_ratio 0.1 \
|
65 |
-
--num_train_epochs 5 \
|
66 |
-
--per_device_train_batch_size 32 \
|
67 |
-
--gradient_accumulation_steps 4 \
|
68 |
-
--per_device_eval_batch_size 32 \
|
69 |
-
--dataloader_num_workers 1 \
|
70 |
-
--logging_strategy steps \
|
71 |
-
--logging_steps 10 \
|
72 |
-
--evaluation_strategy epoch \
|
73 |
-
--save_strategy epoch \
|
74 |
-
--load_best_model_at_end True \
|
75 |
-
--metric_for_best_model accuracy \
|
76 |
-
--save_total_limit 3 \
|
77 |
-
--seed 0
|
78 |
-
'''
|
79 |
|
80 |
### Training hyperparameters
|
81 |
|
@@ -96,10 +54,10 @@ The following hyperparameters were used during training:
|
|
96 |
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|
97 |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
|
98 |
| 0.6089 | 1.0 | 29 | 0.1385 | 0.9962 |
|
99 |
-
| 0.
|
100 |
-
| 0.0835 | 3.0 | 87 | 0.
|
101 |
-
| 0.
|
102 |
-
| 0.
|
103 |
|
104 |
|
105 |
### Framework versions
|
|
|
21 |
- Loss: 0.1385
|
22 |
- Accuracy: 0.9962
|
23 |
|
24 |
+
## Model description
|
|
|
|
|
|
|
25 |
|
26 |
+
More information needed
|
27 |
|
28 |
+
## Intended uses & limitations
|
29 |
+
|
30 |
+
More information needed
|
31 |
+
|
32 |
+
## Training and evaluation data
|
33 |
+
|
34 |
+
More information needed
|
35 |
|
36 |
## Training procedure
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
|
38 |
### Training hyperparameters
|
39 |
|
|
|
54 |
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|
55 |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
|
56 |
| 0.6089 | 1.0 | 29 | 0.1385 | 0.9962 |
|
57 |
+
| 0.1289 | 2.0 | 58 | 0.0510 | 0.9962 |
|
58 |
+
| 0.0835 | 3.0 | 87 | 0.0433 | 0.9885 |
|
59 |
+
| 0.0605 | 4.0 | 116 | 0.0330 | 0.9923 |
|
60 |
+
| 0.0479 | 5.0 | 145 | 0.0273 | 0.9904 |
|
61 |
|
62 |
|
63 |
### Framework versions
|