End of training

README.md

@@ -114,7 +114,7 @@ wandb_log_model:
 
 gradient_accumulation_steps: 2
 micro_batch_size: 2
-num_epochs: 1
+num_epochs: 2
 optimizer: adamw_8bit
 lr_scheduler: cosine
 learning_rate: 0.0002
@@ -154,7 +154,7 @@ special_tokens:
 
 This model is a fine-tuned version of [meta-llama/Llama-3.2-3B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct) on the minpeter/xlam-function-calling-60k-hermes, the minpeter/xlam-irrelevance-7.5k-qwen2.5-72b-distill-hermes, the minpeter/hermes-function-calling-v1-jsonl and the minpeter/hermes-function-calling-v1-jsonl datasets.
 It achieves the following results on the evaluation set:
-- Loss: 0.3492
+- Loss: 0.3335
 
 ## Model description
 
@@ -182,16 +182,19 @@ The following hyperparameters were used during training:
 - optimizer: Use OptimizerNames.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
-- num_epochs: 1.0
+- num_epochs: 2.0
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
 | 0.5354        | 0.0039 | 1    | 0.7727          |
-| 0.4681        | 0.3327 | 85   | 0.3746          |
-| 0.1909        | 0.6654 | 170  | 0.3528          |
-| 0.6071        | 0.9980 | 255  | 0.3492          |
+| 0.4667        | 0.3327 | 85   | 0.3745          |
+| 0.1858        | 0.6654 | 170  | 0.3515          |
+| 0.5982        | 0.9980 | 255  | 0.3440          |
+| 0.1452        | 1.3288 | 340  | 0.3389          |
+| 0.2287        | 1.6614 | 425  | 0.3344          |
+| 0.1441        | 1.9941 | 510  | 0.3335          |
 
 
 ### Framework versions