End of training

README.md

@@ -60,7 +60,7 @@ resume_from_checkpoint:
 logging_steps: 1
 flash_attention: true
 
-warmup_steps: 10
+warmup_steps: 1
 evals_per_epoch: 2
 saves_per_epoch: 1
 weight_decay: 0.0
@@ -75,7 +75,7 @@ special_tokens:
 
 This model is a fine-tuned version of [meta-llama/Llama-3.2-1B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct) on the syvai/no-emotion-reasoning dataset.
 It achieves the following results on the evaluation set:
-- Loss: 1.9291
+- Loss: 0.5303
 
 ## Model description
 
@@ -102,7 +102,7 @@ The following hyperparameters were used during training:
 - total_train_batch_size: 16
 - optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
 - lr_scheduler_type: cosine
-- lr_scheduler_warmup_steps: 10
+- lr_scheduler_warmup_steps: 2
 - num_epochs: 1.0
 
 ### Training results
@@ -110,7 +110,7 @@ The following hyperparameters were used during training:
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
 | 8.4068        | 0.1013 | 1    | 8.7623          |
-| 3.8505        | 0.5063 | 5    | 1.9291          |
+| 0.8452        | 0.5063 | 5    | 0.5303          |
 
 
 ### Framework versions