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: 2
 mlflow_experiment_name: /tmp/4521bc191b2ab6e0_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: 27f3654a-bcca-42f5-8aa3-5e8cec20f5ac
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 27f3654a-bcca-42f5-8aa3-5e8cec20f5ac
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # b15857f8-a87f-8b0e-4d5c-9a19dd87f016
 
 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: 0.9453
 
 ## 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        | 0.0000 | 1    | 1.1955          |
+| 1.1851        | 0.0003 | 10   | 1.1461          |
+| 1.0061        | 0.0006 | 20   | 1.0128          |
+| 0.9271        | 0.0009 | 30   | 0.9618          |
+| 0.8849        | 0.0012 | 40   | 0.9478          |
+| 0.9729        | 0.0015 | 50   | 0.9453          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b5556ca1191bcf4bd88afd4815784f2690e4d25b9bb114ca8466ca6a76908a88
+oid sha256:791746f52d928eca5f9b61d899c4044303f960319b2ef7c003864e24d0d5bdc2
 size 80115210