Hugging Face's logo

thejosango
/
nuha-ajp-binary

Text Classification
Transformers
PyTorch
Safetensors
Arabic
bert
hate-speech
gender-based-violence
arabic
binary-classification
pilot
Eval Results (legacy)
text-embeddings-inference
Model card Files
xet
 Community
4
yazansh commited on
Commit
4964c88
·
1 Parent(s): f88e36e

binary-13

Browse files
Files changed (6) hide show
  1. README.md +18 -18
  2. config.json +1 -1
  3. config.toml +3 -3
  4. pytorch_model.bin +2 -2
  5. tokenizer.json +1 -6
  6. training_args.bin +1 -1
README.md CHANGED
@@ -23,13 +23,13 @@ model-index:
23
  metrics:
24
  - name: F1
25
  type: f1
26
- value: 0.6569430569430569
27
  - name: Precision
28
  type: precision
29
- value: 0.5546558704453441
30
  - name: Recall
31
  type: recall
32
- value: 0.8054875061244487
33
  ---
34
 
35
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
@@ -39,10 +39,10 @@ should probably proofread and complete it, then remove this comment. -->
39
 
40
  This model is a fine-tuned version of [aubmindlab/bert-base-arabertv02-twitter](https://huggingface.co/aubmindlab/bert-base-arabertv02-twitter) on the nuha-dataset dataset.
41
  It achieves the following results on the evaluation set:
42
- - Loss: 1.3847
43
- - F1: 0.6569
44
- - Precision: 0.5547
45
- - Recall: 0.8055
46
  - Support: None
47
 
48
  ## Model description
@@ -67,7 +67,7 @@ The following hyperparameters were used during training:
67
  - eval_batch_size: 64
68
  - seed: 42
69
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
70
- - lr_scheduler_type: linear
71
  - lr_scheduler_warmup_steps: 2000.0
72
  - num_epochs: 5
73
  - label_smoothing_factor: 0.1
@@ -76,16 +76,16 @@ The following hyperparameters were used during training:
76
 
77
  | Training Loss | Epoch | Step | Validation Loss | F1 | Precision | Recall | Support |
78
  |:-------------:|:-----:|:----:|:---------------:|:------:|:---------:|:------:|:-------:|
79
- | 6.3184 | 0.49 | 500 | 2.4352 | 0.5147 | 0.4381 | 0.6237 | None |
80
- | 4.073 | 0.98 | 1000 | 1.7487 | 0.5799 | 0.4432 | 0.8388 | None |
81
- | 3.3917 | 1.47 | 1500 | 1.9366 | 0.3586 | 0.7128 | 0.2396 | None |
82
- | 2.6749 | 1.96 | 2000 | 1.7159 | 0.6176 | 0.4893 | 0.8373 | None |
83
- | 2.5173 | 2.45 | 2500 | 1.9496 | 0.2381 | 0.8423 | 0.1387 | None |
84
- | 2.3263 | 2.94 | 3000 | 1.3238 | 0.6466 | 0.5673 | 0.7516 | None |
85
- | 2.066 | 3.43 | 3500 | 1.2226 | 0.6265 | 0.6263 | 0.6267 | None |
86
- | 1.8409 | 3.92 | 4000 | 1.3447 | 0.6557 | 0.5603 | 0.7903 | None |
87
- | 1.7984 | 4.41 | 4500 | 1.2543 | 0.6542 | 0.5885 | 0.7364 | None |
88
- | 1.6764 | 4.9 | 5000 | 1.3847 | 0.6569 | 0.5547 | 0.8055 | None |
89
 
90
 
91
  ### Framework versions
 
23
  metrics:
24
  - name: F1
25
  type: f1
26
+ value: 0.6555940023068051
27
  - name: Precision
28
  type: precision
29
+ value: 0.6194420226678291
30
  - name: Recall
31
  type: recall
32
+ value: 0.6962273395394415
33
  ---
34
 
35
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 
39
 
40
  This model is a fine-tuned version of [aubmindlab/bert-base-arabertv02-twitter](https://huggingface.co/aubmindlab/bert-base-arabertv02-twitter) on the nuha-dataset dataset.
41
  It achieves the following results on the evaluation set:
42
+ - Loss: 0.5987
43
+ - F1: 0.6556
44
+ - Precision: 0.6194
45
+ - Recall: 0.6962
46
  - Support: None
47
 
48
  ## Model description
 
67
  - eval_batch_size: 64
68
  - seed: 42
69
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
70
+ - lr_scheduler_type: constant_with_warmup
71
  - lr_scheduler_warmup_steps: 2000.0
72
  - num_epochs: 5
73
  - label_smoothing_factor: 0.1
 
76
 
77
  | Training Loss | Epoch | Step | Validation Loss | F1 | Precision | Recall | Support |
78
  |:-------------:|:-----:|:----:|:---------------:|:------:|:---------:|:------:|:-------:|
79
+ | 2.5085 | 0.49 | 500 | 1.0583 | 0.5122 | 0.5269 | 0.4983 | None |
80
+ | 1.5593 | 0.98 | 1000 | 0.7585 | 0.6006 | 0.4928 | 0.7687 | None |
81
+ | 1.198 | 1.47 | 1500 | 0.6285 | 0.5762 | 0.6121 | 0.5443 | None |
82
+ | 0.984 | 1.96 | 2000 | 0.6197 | 0.6279 | 0.5648 | 0.7070 | None |
83
+ | 0.9388 | 2.45 | 2500 | 0.5948 | 0.5851 | 0.6627 | 0.5238 | None |
84
+ | 0.8417 | 2.94 | 3000 | 0.6793 | 0.6428 | 0.5259 | 0.8266 | None |
85
+ | 0.804 | 3.43 | 3500 | 0.5996 | 0.6548 | 0.5716 | 0.7663 | None |
86
+ | 0.7546 | 3.92 | 4000 | 0.6484 | 0.6584 | 0.5763 | 0.7678 | None |
87
+ | 0.7425 | 4.41 | 4500 | 0.5629 | 0.6503 | 0.6181 | 0.6859 | None |
88
+ | 0.712 | 4.9 | 5000 | 0.5987 | 0.6556 | 0.6194 | 0.6962 | None |
89
 
90
 
91
  ### Framework versions
config.json CHANGED
@@ -23,7 +23,7 @@
23
  "max_position_embeddings": 512,
24
  "model_type": "bert",
25
  "num_attention_heads": 12,
26
- "num_hidden_layers": 12,
27
  "pad_token_id": 0,
28
  "position_embedding_type": "absolute",
29
  "problem_type": "single_label_classification",
 
23
  "max_position_embeddings": 512,
24
  "model_type": "bert",
25
  "num_attention_heads": 12,
26
+ "num_hidden_layers": 6,
27
  "pad_token_id": 0,
28
  "position_embedding_type": "absolute",
29
  "problem_type": "single_label_classification",
config.toml CHANGED
@@ -1,5 +1,5 @@
1
  [experiment]
2
- name = "binary-12"
3
  type = "binary"
4
 
5
 
@@ -16,7 +16,7 @@ revision = "main"
16
  hidden_dropout_prob = 0.3
17
  attention_probs_dropout_prob = 0.3
18
  classifier_dropout = 0.3
19
- #num_hidden_layers = 6
20
  #num_attention_heads = 12
21
  #hidden_size = 768
22
  #intermediate_size= null
@@ -25,7 +25,7 @@ classifier_dropout = 0.3
25
  [training]
26
  num_train_epochs = 5
27
  warmup_steps = 2e3
28
- lr_scheduler_type = "linear"
29
  learning_rate = 3e-5
30
  per_device_train_batch_size = 32
31
  per_device_eval_batch_size = 64
 
1
  [experiment]
2
+ name = "binary-13"
3
  type = "binary"
4
 
5
 
 
16
  hidden_dropout_prob = 0.3
17
  attention_probs_dropout_prob = 0.3
18
  classifier_dropout = 0.3
19
+ num_hidden_layers = 6
20
  #num_attention_heads = 12
21
  #hidden_size = 768
22
  #intermediate_size= null
 
25
  [training]
26
  num_train_epochs = 5
27
  warmup_steps = 2e3
28
+ lr_scheduler_type = "constant_with_warmup"
29
  learning_rate = 3e-5
30
  per_device_train_batch_size = 32
31
  per_device_eval_batch_size = 64
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:07c3883e8dda6062e40847d5a899d48a7b0efca48ab28a910f40992a01964619
3
- size 540847921
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:faa1c66eb99660799297ea153e0e9e220570a7f4cfecb2788c95651352c486bc
3
+ size 370706033
tokenizer.json CHANGED
@@ -1,11 +1,6 @@
1
  {
2
  "version": "1.0",
3
- "truncation": {
4
- "direction": "Right",
5
- "max_length": 512,
6
- "strategy": "LongestFirst",
7
- "stride": 0
8
- },
9
  "padding": null,
10
  "added_tokens": [
11
  {
 
1
  {
2
  "version": "1.0",
3
+ "truncation": null,
 
 
 
 
 
4
  "padding": null,
5
  "added_tokens": [
6
  {
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a4f9dd429a6f3e5ca170fd340f940feee05a4e084a788ec1b83942e6396544be
3
  size 4091
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f764a8af81060b173458702e69fbe2eb16017942aa78b88adb5be4f2f5e0ea9b
3
  size 4091