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/75e2dfb45725a08d_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: decda85a-cc80-403f-8f61-2d4405578f06
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: decda85a-cc80-403f-8f61-2d4405578f06
-warmup_steps: 1
+warmup_steps: 2
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # decda85a-cc80-403f-8f61-2d4405578f06
 
 This model is a fine-tuned version of [unsloth/Qwen2.5-Math-1.5B](https://huggingface.co/unsloth/Qwen2.5-Math-1.5B) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: nan
 
 ## Model description
 
@@ -129,13 +131,17 @@ 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: 14
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
 | No log        | 0.0748 | 1    | nan             |
+| No log        | 0.2243 | 3    | nan             |
+| No log        | 0.4486 | 6    | nan             |
+| No log        | 0.6729 | 9    | nan             |
+| 0.0           | 0.8972 | 12   | nan             |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e9f24aa57534c8274ddbef0bd13f8077475839d193b405924fb929802a9009f6
+oid sha256:68b9c025307d8c0f82722c87b8783cc11fd6b9ad4b3b10e979d23d553ea78723
 size 37070634