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/6328008f853d87e9_train_data.json
 model_type: AutoModelForCausalLM
@@ -93,7 +93,7 @@ wandb_name: e1af209e-9139-4d0e-a422-fbcdda67f259
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: e1af209e-9139-4d0e-a422-fbcdda67f259
-warmup_steps: 1
+warmup_steps: 2
 weight_decay: 0.0
 xformers_attention: null
 
@@ -104,6 +104,8 @@ xformers_attention: null
 # e1af209e-9139-4d0e-a422-fbcdda67f259
 
 This model is a fine-tuned version of [JackFram/llama-160m](https://huggingface.co/JackFram/llama-160m) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 3.9678
 
 ## 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        | 0.0167 | 1    | 6.8287          |
+| 6.479         | 0.1667 | 10   | 6.0251          |
+| 5.4595        | 0.3333 | 20   | 5.0056          |
+| 4.5889        | 0.5    | 30   | 4.3424          |
+| 4.0775        | 0.6667 | 40   | 4.0246          |
+| 3.8715        | 0.8333 | 50   | 3.9678          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6a374ba8a6a1f03c2ba66519e0bdf1e79f4533a09454b8dc5ee5e32cfc5c4bd4
+oid sha256:3348710547a7ae9e5f57741ac3c320cdeeda02d75995ef6466c94da42ba329ea
 size 6843306