End of training

README.md

@@ -65,7 +65,7 @@ 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/365ee670b7407e0c_train_data.json
 model_type: AutoModelForCausalLM
@@ -92,7 +92,7 @@ wandb_name: 552a5907-3f1e-4ce4-85f7-8c59cdfc4191
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 552a5907-3f1e-4ce4-85f7-8c59cdfc4191
-warmup_steps: 1
+warmup_steps: 2
 weight_decay: 0.0
 xformers_attention: null
 
@@ -103,6 +103,8 @@ xformers_attention: null
 # 552a5907-3f1e-4ce4-85f7-8c59cdfc4191
 
 This model is a fine-tuned version of [Korabbit/llama-2-ko-7b](https://huggingface.co/Korabbit/llama-2-ko-7b) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0576
 
 ## Model description
 
@@ -130,13 +132,18 @@ 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        | 0.0003 | 1    | 2.6466          |
+| 1.7919        | 0.0034 | 10   | 0.4523          |
+| 0.1707        | 0.0068 | 20   | 0.0758          |
+| 0.0686        | 0.0103 | 30   | 0.0604          |
+| 0.0596        | 0.0137 | 40   | 0.0583          |
+| 0.0563        | 0.0171 | 50   | 0.0576          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dd4d3209e15b8900919f38ccc8d161f287fc2dcc457fc5420ff76a427cc22c7d
+oid sha256:e268061a8945f83b8be38776945537b16df4901260466be11576a20fad278595
 size 80115210