End of training

README.md

@@ -66,7 +66,7 @@ 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/6d57673c12aeff10_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: 76155263-4121-4da4-a8d5-5cfb27f05540
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 76155263-4121-4da4-a8d5-5cfb27f05540
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # 6197c78b-29de-4067-a6b6-bfc99b03b4f4
 
 This model is a fine-tuned version of [unsloth/llama-2-7b](https://huggingface.co/unsloth/llama-2-7b) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.8988
 
 ## 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    | 1.1126          |
+| 0.9738        | 0.0003 | 10   | 1.0696          |
+| 0.9683        | 0.0007 | 20   | 0.9722          |
+| 0.9272        | 0.0010 | 30   | 0.9234          |
+| 0.8495        | 0.0013 | 40   | 0.9026          |
+| 0.8792        | 0.0017 | 50   | 0.8988          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d239c67e6b0187ef5e1e8c993c4d3d4233da2f0a4781701429acb4dcbda7a5c3
+oid sha256:8b4ef719fe4383303b862c799f8326ba1e9a79320e2d70f0fc0da58d6425b897
 size 80115210