mlm-3

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: 4.7143
+- Loss: 4.2591
 
 ## Model description
 
@@ -35,29 +35,26 @@ More information needed
 ### Training hyperparameters
 
 The following hyperparameters were used during training:
-- learning_rate: 5e-05
-- train_batch_size: 16
+- learning_rate: 0.0001
+- train_batch_size: 32
 - eval_batch_size: 32
 - seed: 42
-- gradient_accumulation_steps: 4
+- 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: 1000.0
-- num_epochs: 5
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 2000.0
+- num_epochs: 3
 - label_smoothing_factor: 0.1
 
 ### Training results
 
 | Training Loss | Epoch | Step | Validation Loss |
 |:-------------:|:-----:|:----:|:---------------:|
-| 5.6551        | 0.64  | 500  | 5.3465          |
-| 5.2489        | 1.28  | 1000 | 5.1498          |
-| 5.1004        | 1.91  | 1500 | 4.9168          |
-| 4.9289        | 2.55  | 2000 | 4.9202          |
-| 4.9081        | 3.19  | 2500 | 4.8600          |
-| 4.7641        | 3.83  | 3000 | 4.7688          |
-| 4.6543        | 4.47  | 3500 | 4.7143          |
+| 4.6174        | 0.64  | 500  | 4.4343          |
+| 4.3509        | 1.28  | 1000 | 4.2799          |
+| 4.3392        | 1.91  | 1500 | 4.2559          |
+| 4.2514        | 2.55  | 2000 | 4.2591          |
 
 
 ### Framework versions

config.json

@@ -15,7 +15,7 @@
   "max_position_embeddings": 512,
   "model_type": "bert",
   "num_attention_heads": 12,
-  "num_hidden_layers": 8,
+  "num_hidden_layers": 12,
   "pad_token_id": 0,
   "position_embedding_type": "absolute",
   "torch_dtype": "float32",

config.toml

@@ -1,5 +1,5 @@
 [experiment]
-name = "mlm-2"
+name = "mlm-3"
 type = "mlm"
 
 
@@ -16,20 +16,20 @@ revision = "main"
 #hidden_dropout_prob = 0.2
 #attention_probs_dropout_prob = 0.2
 #classifier_dropout = 0.2
-num_hidden_layers = 8 
+#num_hidden_layers = 8 
 #num_attention_heads = 12
 #hidden_size = 768
 #intermediate_size= null
 
 
 [training]
-num_train_epochs = 5
-warmup_steps = 1e3
-lr_scheduler_type = "constant"
-learning_rate = 5e-5
-per_device_train_batch_size = 16 
+num_train_epochs = 3
+warmup_steps = 2e3
+lr_scheduler_type = "linear"
+learning_rate = 1e-4
+per_device_train_batch_size = 32 
 per_device_eval_batch_size = 32
-gradient_accumulation_steps = 4
+gradient_accumulation_steps = 2
 weight_decay = 0.01
 label_smoothing_factor = 0.1
 weighted_loss = false

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:77ba656ca0a468c5254acd0f409bf4736a1cbf0140d9064fd0e21c59cbe176a3
-size 427676597
+oid sha256:812c36a44c7713e58775c0aac94aa7b15c6509792d3c97038a7e4651c3cf8d29
+size 541104501

tokenizer.json

@@ -1,19 +1,7 @@
 {
   "version": "1.0",
-  "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]"
-  },
+  "truncation": null,
+  "padding": null,
   "added_tokens": [
     {
       "id": 0,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:14b7b9084803fcab4baf4a0bb822aa03dd0f81f6311c38ae7f3411489a2dd872
+oid sha256:d94db48e3cc2b831068369c7fecfafa07c51db5aaed7a2f6a9959a46edebe263
 size 4027