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/cb900b6d3c8d04fe_train_data.json
 model_type: AutoModelForCausalLM
@@ -93,7 +93,7 @@ wandb_name: e514f4c6-15bd-4743-8388-5f183bc3f4fb
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: e514f4c6-15bd-4743-8388-5f183bc3f4fb
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -104,6 +104,8 @@ xformers_attention: null
 # e514f4c6-15bd-4743-8388-5f183bc3f4fb
 
 This model is a fine-tuned version of [echarlaix/tiny-random-mistral](https://huggingface.co/echarlaix/tiny-random-mistral) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 10.3779
 
 ## Model description
 
@@ -130,14 +132,19 @@ 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        | 0.0001 | 1    | 10.3783         |
+| 41.4942       | 0.0008 | 10   | 10.3782         |
+| 41.5298       | 0.0015 | 20   | 10.3781         |
+| 41.5211       | 0.0023 | 30   | 10.3780         |
+| 41.5074       | 0.0031 | 40   | 10.3779         |
+| 41.5059       | 0.0038 | 50   | 10.3779         |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e26afc21c44fa63fea2a5a794ea34344ce177fde6575b3addc0ec63391b536eb
+oid sha256:3f0ebae252372b9bea400fbc77c25a840e7934bd5d93c9dbbc1c14e5101eb355
 size 37506