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: 8
 mlflow_experiment_name: /tmp/cdaca08d5c55e915_train_data.json
 model_type: AutoModelForCausalLM
@@ -93,7 +93,7 @@ wandb_name: 586d61d7-7f61-46e4-801f-f3c4abb10e52
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 586d61d7-7f61-46e4-801f-f3c4abb10e52
-warmup_steps: 1
+warmup_steps: 2
 weight_decay: 0.0
 xformers_attention: null
 
@@ -104,6 +104,8 @@ xformers_attention: null
 # 586d61d7-7f61-46e4-801f-f3c4abb10e52
 
 This model is a fine-tuned version of [JackFram/llama-68m](https://huggingface.co/JackFram/llama-68m) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 1.5735
 
 ## Model description
 
@@ -131,13 +133,18 @@ 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        | 0.0143 | 1    | 2.3939          |
+| 2.3322        | 0.1435 | 10   | 2.1275          |
+| 1.9772        | 0.2870 | 20   | 1.8150          |
+| 1.7502        | 0.4305 | 30   | 1.6535          |
+| 1.6222        | 0.5740 | 40   | 1.5852          |
+| 1.5933        | 0.7175 | 50   | 1.5735          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:19936e624e5b196e0528f557bc1a56aa0ebf8a033ca7ce1ce65cdac736c1413e
+oid sha256:3aaa7fdb68388eb5e59d33518820be4cadbc4afe70c4f5de96aa53c731f07358
 size 1140674