mlm-5

README.md

@@ -16,7 +16,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 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.
 It achieves the following results on the evaluation set:
-- Loss: 4.3014
+- Loss: 4.1000
 
 ## Model description
 
@@ -35,23 +35,33 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 1e-05
-- train_batch_size: 64
-- eval_batch_size: 64
+- learning_rate: 5e-05
+- train_batch_size: 32
+- eval_batch_size: 32
 - seed: 42
+- gradient_accumulation_steps: 2
+- total_train_batch_size: 64
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: constant
-- lr_scheduler_warmup_steps: 500.0
-- num_epochs: 1
+- lr_scheduler_warmup_steps: 1000.0
+- num_epochs: 3
 - label_smoothing_factor: 0.1
 
 ### Training results
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| 4.6458        | 0.25  | 500  | 4.4435          |
-| 4.4927        | 0.5   | 1000 | 4.3572          |
-| 4.4381        | 0.75  | 1500 | 4.3014          |
+| 4.5289        | 0.25  | 500  | 4.3484          |
+| 4.4276        | 0.5   | 1000 | 4.2974          |
+| 4.402         | 0.75  | 1500 | 4.2336          |
+| 4.3494        | 1.0   | 2000 | 4.1649          |
+| 4.2526        | 1.26  | 2500 | 4.1969          |
+| 4.2335        | 1.51  | 3000 | 4.1577          |
+| 4.2401        | 1.76  | 3500 | 4.1136          |
+| 4.1889        | 2.01  | 4000 | 4.1469          |
+| 4.1366        | 2.26  | 4500 | 4.1297          |
+| 4.1435        | 2.51  | 5000 | 4.0907          |
+| 4.1207        | 2.76  | 5500 | 4.1000          |
 
 
 ### Framework versions

config.toml

@@ -1,5 +1,5 @@
 [experiment]
-name = "mlm-4"
+name = "mlm-5"
 type = "mlm"
 
 
@@ -12,18 +12,17 @@ undersampling_strategy = false
 
 [model]
 pretrained_model_name_or_path = "aubmindlab/bert-base-arabertv02-twitter"
-revision = "main"
 
 
 [training]
-num_train_epochs = 1 
-warmup_steps = 5e2
+num_train_epochs = 3 
+warmup_steps = 1e3
 lr_scheduler_type = "constant"
-learning_rate = 1e-5
-per_device_train_batch_size = 64 
-per_device_eval_batch_size = 64
-gradient_accumulation_steps = 1 
-weight_decay = 0.01
+learning_rate = 5e-5
+per_device_train_batch_size = 32
+per_device_eval_batch_size = 32
+gradient_accumulation_steps = 2
+weight_decay = 0.00
 label_smoothing_factor = 0.1
 weighted_loss = false
 early_stopping_patience = 5

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4606e079bff12241fab620c6ea27ccf1c6a4a95ad320b348353b82e869dad9ee
+oid sha256:7aed42dcd14312adec1767d61596c1213bfdac1ca5e6620ff838b7c82551b6a2
 size 541104501

tokenizer.json

@@ -1,7 +1,19 @@
 {
   "version": "1.0",
-  "truncation": null,
-  "padding": null,
+  "truncation": {
+    "direction": "Right",
+    "max_length": 512,
+    "strategy": "LongestFirst",
+    "stride": 0
+  },
+  "padding": {
+    "strategy": "BatchLongest",
+    "direction": "Right",
+    "pad_to_multiple_of": null,
+    "pad_id": 0,
+    "pad_type_id": 0,
+    "pad_token": "[PAD]"
+  },
   "added_tokens": [
     {
       "id": 0,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6420bbe060e364db418484f88662d36095b1bb69ce021b673848354839c0617
+oid sha256:591a687be7219249edf221ee703b4450d76d95a56bc77b37a6061088d416e8a0
 size 4091