End of training

README.md

@@ -66,7 +66,7 @@ lora_model_dir: null
 lora_r: 8
 lora_target_linear: true
 lr_scheduler: cosine
-max_steps: 1
+max_steps: 50
 micro_batch_size: 2
 mlflow_experiment_name: /tmp/dc5b95a08dd82e2f_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: 171b38ed-2620-4b56-837a-a1fd81b5e669
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 171b38ed-2620-4b56-837a-a1fd81b5e669
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # 171b38ed-2620-4b56-837a-a1fd81b5e669
 
 This model is a fine-tuned version of [unsloth/codegemma-7b](https://huggingface.co/unsloth/codegemma-7b) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0137
 
 ## Model description
 
@@ -128,14 +130,19 @@ The following hyperparameters were used during training:
 - total_train_batch_size: 8
 - optimizer: Use OptimizerNames.ADAMW_BNB 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: 2
-- training_steps: 1
+- lr_scheduler_warmup_steps: 10
+- training_steps: 50
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
 | No log        | 0.0002 | 1    | 7.1835          |
+| 5.2795        | 0.0024 | 10   | 3.5841          |
+| 0.5604        | 0.0049 | 20   | 0.1417          |
+| 0.0513        | 0.0073 | 30   | 0.0460          |
+| 0.0019        | 0.0098 | 40   | 0.0192          |
+| 0.0188        | 0.0122 | 50   | 0.0137          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:acf4958f80d5f1999b3569aa3337d3bf1269601c842f2e41a6af3b9e173fae62
+oid sha256:459e135b808cc24fb856e0b5ca7d14e7d3a7123bf73261b0eef7bced0b392e9d
 size 100149034