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/4c1a92c81a339635_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: c640ee77-d816-4ae9-8f2b-f50c8a84197c
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: c640ee77-d816-4ae9-8f2b-f50c8a84197c
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # 6ed444f9-942a-64a6-1370-d45dd99ca881
 
 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: 1.5479
 
 ## 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        | 0.0017 | 1    | 2.3447          |
+| 2.2288        | 0.0174 | 10   | 2.0354          |
+| 1.7169        | 0.0348 | 20   | 1.6828          |
+| 1.6979        | 0.0521 | 30   | 1.5826          |
+| 1.5266        | 0.0695 | 40   | 1.5529          |
+| 1.5404        | 0.0869 | 50   | 1.5479          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:72a5730ad3887cc36ab0b2458d4f1a490bd0a516e1a4a16de6ddbb7a92e4860c
+oid sha256:93994e5c747bc034b40fd6b0f60c6c07b6301f8fbd07b08e85846de2c72c37bd
 size 80115210