End of training

README.md

@@ -75,7 +75,7 @@ lora_target_modules:
 lr_scheduler: cosine
 max_memory:
   0: 70GiB
-max_steps: 5
+max_steps: 50
 micro_batch_size: 2
 mlflow_experiment_name: /tmp/1c2d230e0db4aaf5_train_data.json
 model_type: AutoModelForCausalLM
@@ -118,7 +118,7 @@ xformers_attention: null
 
 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: 0.8985
+- Loss: 0.0540
 
 ## Model description
 
@@ -146,15 +146,16 @@ 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: 10
-- training_steps: 5
+- training_steps: 50
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
 | 0.9265        | 0.0024 | 1    | 0.9870          |
-| 1.0298        | 0.0048 | 2    | 0.9388          |
-| 0.9506        | 0.0095 | 4    | 0.8985          |
+| 0.3231        | 0.0310 | 13   | 0.3032          |
+| 0.1219        | 0.0620 | 26   | 0.0904          |
+| 0.0602        | 0.0930 | 39   | 0.0540          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1508bbce101c2fb481eb76fc83c844edf9e2118f937bbb6efa6c399b6095ed49
+oid sha256:a5114fa34ed382549acc5ff56201c6cc1710b082b8968d1e0b63ff44e58ca4d0
 size 80115210