End of training

README.md

@@ -64,7 +64,7 @@ lora_model_dir: null
 lora_r: 8
 lora_target_linear: true
 lr_scheduler: cosine
-max_steps: 1
+max_steps: 50
 micro_batch_size: 8
 mlflow_experiment_name: /tmp/76e846cebbc78c1f_train_data.json
 model_type: AutoModelForCausalLM
@@ -89,7 +89,7 @@ wandb_name: 3dda6899-9e6e-4b50-964b-f63dd87f4fab
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 3dda6899-9e6e-4b50-964b-f63dd87f4fab
-warmup_steps: 1
+warmup_steps: 2
 weight_decay: 0.0
 xformers_attention: null
 
@@ -100,6 +100,8 @@ xformers_attention: null
 # 3dda6899-9e6e-4b50-964b-f63dd87f4fab
 
 This model is a fine-tuned version of [oopsung/llama2-7b-koNqa-test-v1](https://huggingface.co/oopsung/llama2-7b-koNqa-test-v1) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.9005
 
 ## Model description
 
@@ -127,13 +129,18 @@ The following hyperparameters were used during training:
 - 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
+- training_steps: 50
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
 | No log        | 0.0006 | 1    | 2.7885          |
+| 2.5126        | 0.0061 | 10   | 1.9133          |
+| 1.4967        | 0.0121 | 20   | 1.1760          |
+| 1.0583        | 0.0182 | 30   | 0.9798          |
+| 0.9362        | 0.0243 | 40   | 0.9117          |
+| 0.9095        | 0.0304 | 50   | 0.9005          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ccc40e5d0b773dae81fdfaa9486eae8ddee262237effe231333f3db1971e08f8
+oid sha256:7dd6ff98a693e0f62fb7480e3f62108234aceddc050b989f2289ce177f3c9209
 size 80115210