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/ee570cc5bb9ff625_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: 829b1f08-58c3-4952-aa0e-273d0559bc1f
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 829b1f08-58c3-4952-aa0e-273d0559bc1f
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # 5b08da3b-fdc8-4b36-f59d-55947f9df3d1
 
 This model is a fine-tuned version of [unsloth/codellama-7b](https://huggingface.co/unsloth/codellama-7b) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.4485
 
 ## 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        | 0.0001 | 1    | 0.7562          |
+| 0.7311        | 0.0006 | 10   | 0.6244          |
+| 0.516         | 0.0012 | 20   | 0.4802          |
+| 0.46          | 0.0018 | 30   | 0.4582          |
+| 0.4092        | 0.0023 | 40   | 0.4501          |
+| 0.4351        | 0.0029 | 50   | 0.4485          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4e12beb3b17805010442ea5c733060ac232decf66f775196a2da536bc01416c3
+oid sha256:abb96a284aa65e73fa84b0461772bc2a9c7aa827325a40bccce6a60e5a534596
 size 80115210