abdouaziiz
commited on
Commit
•
48f52fe
1
Parent(s):
7455a7d
End of training
Browse files- README.md +16 -4
- all_results.json +15 -0
- eval_results.json +9 -0
- train_results.json +9 -0
- trainer_state.json +166 -0
README.md
CHANGED
@@ -3,11 +3,23 @@ license: apache-2.0
|
|
3 |
base_model: openai/whisper-tiny
|
4 |
tags:
|
5 |
- generated_from_trainer
|
|
|
|
|
6 |
metrics:
|
7 |
- wer
|
8 |
model-index:
|
9 |
- name: whisper-ling-asr
|
10 |
-
results:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
---
|
12 |
|
13 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
@@ -16,10 +28,10 @@ should probably proofread and complete it, then remove this comment. -->
|
|
16 |
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/abdouaziz/wolof-asr/runs/4gbjp7dq)
|
17 |
# whisper-ling-asr
|
18 |
|
19 |
-
This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on
|
20 |
It achieves the following results on the evaluation set:
|
21 |
-
- Loss: 0.
|
22 |
-
- Wer: 0.
|
23 |
|
24 |
## Model description
|
25 |
|
|
|
3 |
base_model: openai/whisper-tiny
|
4 |
tags:
|
5 |
- generated_from_trainer
|
6 |
+
datasets:
|
7 |
+
- KasuleTrevor/lingala_20hr
|
8 |
metrics:
|
9 |
- wer
|
10 |
model-index:
|
11 |
- name: whisper-ling-asr
|
12 |
+
results:
|
13 |
+
- task:
|
14 |
+
name: Automatic Speech Recognition
|
15 |
+
type: automatic-speech-recognition
|
16 |
+
dataset:
|
17 |
+
name: KasuleTrevor/lingala_20hr
|
18 |
+
type: KasuleTrevor/lingala_20hr
|
19 |
+
metrics:
|
20 |
+
- name: Wer
|
21 |
+
type: wer
|
22 |
+
value: 0.2979101583539804
|
23 |
---
|
24 |
|
25 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
28 |
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/abdouaziz/wolof-asr/runs/4gbjp7dq)
|
29 |
# whisper-ling-asr
|
30 |
|
31 |
+
This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the KasuleTrevor/lingala_20hr dataset.
|
32 |
It achieves the following results on the evaluation set:
|
33 |
+
- Loss: 0.4835
|
34 |
+
- Wer: 0.2979
|
35 |
|
36 |
## Model description
|
37 |
|
all_results.json
ADDED
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 14.925373134328359,
|
3 |
+
"eval_loss": 0.4834844172000885,
|
4 |
+
"eval_runtime": 473.9304,
|
5 |
+
"eval_samples": 1099,
|
6 |
+
"eval_samples_per_second": 2.319,
|
7 |
+
"eval_steps_per_second": 0.146,
|
8 |
+
"eval_wer": 0.2979101583539804,
|
9 |
+
"total_flos": 1.96882148081664e+18,
|
10 |
+
"train_loss": 0.1769935465335846,
|
11 |
+
"train_runtime": 17207.4544,
|
12 |
+
"train_samples": 5358,
|
13 |
+
"train_samples_per_second": 11.158,
|
14 |
+
"train_steps_per_second": 0.697
|
15 |
+
}
|
eval_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 14.925373134328359,
|
3 |
+
"eval_loss": 0.4834844172000885,
|
4 |
+
"eval_runtime": 473.9304,
|
5 |
+
"eval_samples": 1099,
|
6 |
+
"eval_samples_per_second": 2.319,
|
7 |
+
"eval_steps_per_second": 0.146,
|
8 |
+
"eval_wer": 0.2979101583539804
|
9 |
+
}
|
train_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 14.925373134328359,
|
3 |
+
"total_flos": 1.96882148081664e+18,
|
4 |
+
"train_loss": 0.1769935465335846,
|
5 |
+
"train_runtime": 17207.4544,
|
6 |
+
"train_samples": 5358,
|
7 |
+
"train_samples_per_second": 11.158,
|
8 |
+
"train_steps_per_second": 0.697
|
9 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,166 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.4834844172000885,
|
3 |
+
"best_model_checkpoint": "whisper-ling-asr/checkpoint-2000",
|
4 |
+
"epoch": 14.925373134328359,
|
5 |
+
"eval_steps": 1000,
|
6 |
+
"global_step": 5000,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [
|
11 |
+
{
|
12 |
+
"epoch": 1.4925373134328357,
|
13 |
+
"grad_norm": 7.594086647033691,
|
14 |
+
"learning_rate": 9.625941422594142e-06,
|
15 |
+
"loss": 0.8297,
|
16 |
+
"step": 500
|
17 |
+
},
|
18 |
+
{
|
19 |
+
"epoch": 2.9850746268656714,
|
20 |
+
"grad_norm": 8.005925178527832,
|
21 |
+
"learning_rate": 9.207531380753139e-06,
|
22 |
+
"loss": 0.3703,
|
23 |
+
"step": 1000
|
24 |
+
},
|
25 |
+
{
|
26 |
+
"epoch": 2.9850746268656714,
|
27 |
+
"eval_loss": 0.5100315809249878,
|
28 |
+
"eval_runtime": 410.9956,
|
29 |
+
"eval_samples_per_second": 2.674,
|
30 |
+
"eval_steps_per_second": 0.168,
|
31 |
+
"eval_wer": 0.27345685661962726,
|
32 |
+
"step": 1000
|
33 |
+
},
|
34 |
+
{
|
35 |
+
"epoch": 4.477611940298507,
|
36 |
+
"grad_norm": 4.732856273651123,
|
37 |
+
"learning_rate": 8.789121338912134e-06,
|
38 |
+
"loss": 0.2182,
|
39 |
+
"step": 1500
|
40 |
+
},
|
41 |
+
{
|
42 |
+
"epoch": 5.970149253731344,
|
43 |
+
"grad_norm": 4.545703411102295,
|
44 |
+
"learning_rate": 8.37071129707113e-06,
|
45 |
+
"loss": 0.1487,
|
46 |
+
"step": 2000
|
47 |
+
},
|
48 |
+
{
|
49 |
+
"epoch": 5.970149253731344,
|
50 |
+
"eval_loss": 0.4834844172000885,
|
51 |
+
"eval_runtime": 462.8067,
|
52 |
+
"eval_samples_per_second": 2.375,
|
53 |
+
"eval_steps_per_second": 0.149,
|
54 |
+
"eval_wer": 0.2979101583539804,
|
55 |
+
"step": 2000
|
56 |
+
},
|
57 |
+
{
|
58 |
+
"epoch": 7.462686567164179,
|
59 |
+
"grad_norm": 3.4771857261657715,
|
60 |
+
"learning_rate": 7.953138075313808e-06,
|
61 |
+
"loss": 0.0874,
|
62 |
+
"step": 2500
|
63 |
+
},
|
64 |
+
{
|
65 |
+
"epoch": 8.955223880597014,
|
66 |
+
"grad_norm": 2.7059333324432373,
|
67 |
+
"learning_rate": 7.534728033472803e-06,
|
68 |
+
"loss": 0.0556,
|
69 |
+
"step": 3000
|
70 |
+
},
|
71 |
+
{
|
72 |
+
"epoch": 8.955223880597014,
|
73 |
+
"eval_loss": 0.5049698948860168,
|
74 |
+
"eval_runtime": 404.4404,
|
75 |
+
"eval_samples_per_second": 2.717,
|
76 |
+
"eval_steps_per_second": 0.171,
|
77 |
+
"eval_wer": 0.254389744694603,
|
78 |
+
"step": 3000
|
79 |
+
},
|
80 |
+
{
|
81 |
+
"epoch": 10.447761194029852,
|
82 |
+
"grad_norm": 2.1922836303710938,
|
83 |
+
"learning_rate": 7.116317991631799e-06,
|
84 |
+
"loss": 0.0295,
|
85 |
+
"step": 3500
|
86 |
+
},
|
87 |
+
{
|
88 |
+
"epoch": 11.940298507462687,
|
89 |
+
"grad_norm": 0.999194324016571,
|
90 |
+
"learning_rate": 6.697907949790795e-06,
|
91 |
+
"loss": 0.0171,
|
92 |
+
"step": 4000
|
93 |
+
},
|
94 |
+
{
|
95 |
+
"epoch": 11.940298507462687,
|
96 |
+
"eval_loss": 0.5400618314743042,
|
97 |
+
"eval_runtime": 374.2093,
|
98 |
+
"eval_samples_per_second": 2.937,
|
99 |
+
"eval_steps_per_second": 0.184,
|
100 |
+
"eval_wer": 0.23704621350856403,
|
101 |
+
"step": 4000
|
102 |
+
},
|
103 |
+
{
|
104 |
+
"epoch": 13.432835820895523,
|
105 |
+
"grad_norm": 1.5301833152770996,
|
106 |
+
"learning_rate": 6.279497907949791e-06,
|
107 |
+
"loss": 0.0084,
|
108 |
+
"step": 4500
|
109 |
+
},
|
110 |
+
{
|
111 |
+
"epoch": 14.925373134328359,
|
112 |
+
"grad_norm": 0.46232014894485474,
|
113 |
+
"learning_rate": 5.8610878661087865e-06,
|
114 |
+
"loss": 0.0051,
|
115 |
+
"step": 5000
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"epoch": 14.925373134328359,
|
119 |
+
"eval_loss": 0.5673447251319885,
|
120 |
+
"eval_runtime": 355.0078,
|
121 |
+
"eval_samples_per_second": 3.096,
|
122 |
+
"eval_steps_per_second": 0.194,
|
123 |
+
"eval_wer": 0.21765593019498006,
|
124 |
+
"step": 5000
|
125 |
+
},
|
126 |
+
{
|
127 |
+
"epoch": 14.925373134328359,
|
128 |
+
"step": 5000,
|
129 |
+
"total_flos": 1.96882148081664e+18,
|
130 |
+
"train_loss": 0.1769935465335846,
|
131 |
+
"train_runtime": 17207.4544,
|
132 |
+
"train_samples_per_second": 11.158,
|
133 |
+
"train_steps_per_second": 0.697
|
134 |
+
}
|
135 |
+
],
|
136 |
+
"logging_steps": 500,
|
137 |
+
"max_steps": 12000,
|
138 |
+
"num_input_tokens_seen": 0,
|
139 |
+
"num_train_epochs": 36,
|
140 |
+
"save_steps": 1000,
|
141 |
+
"stateful_callbacks": {
|
142 |
+
"EarlyStoppingCallback": {
|
143 |
+
"args": {
|
144 |
+
"early_stopping_patience": 3,
|
145 |
+
"early_stopping_threshold": 0.0
|
146 |
+
},
|
147 |
+
"attributes": {
|
148 |
+
"early_stopping_patience_counter": 0
|
149 |
+
}
|
150 |
+
},
|
151 |
+
"TrainerControl": {
|
152 |
+
"args": {
|
153 |
+
"should_epoch_stop": false,
|
154 |
+
"should_evaluate": false,
|
155 |
+
"should_log": false,
|
156 |
+
"should_save": true,
|
157 |
+
"should_training_stop": true
|
158 |
+
},
|
159 |
+
"attributes": {}
|
160 |
+
}
|
161 |
+
},
|
162 |
+
"total_flos": 1.96882148081664e+18,
|
163 |
+
"train_batch_size": 16,
|
164 |
+
"trial_name": null,
|
165 |
+
"trial_params": null
|
166 |
+
}
|