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/0b44824add214a21_train_data.json
 model_type: AutoModelForCausalLM
@@ -93,7 +93,7 @@ wandb_name: ee18d753-a67f-44fb-bd48-52c33d7962f4
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: ee18d753-a67f-44fb-bd48-52c33d7962f4
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -104,6 +104,8 @@ xformers_attention: null
 # ee18d753-a67f-44fb-bd48-52c33d7962f4
 
 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: 2.8571
 
 ## 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        | 0.0104 | 1    | 3.6499          |
+| 3.4354        | 0.1036 | 10   | 3.3805          |
+| 3.0929        | 0.2073 | 20   | 3.0835          |
+| 2.9533        | 0.3109 | 30   | 2.9467          |
+| 2.863         | 0.4145 | 40   | 2.8700          |
+| 2.8184        | 0.5181 | 50   | 2.8571          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d0d05a491e8d4704df2c13065e7df7fe5db707e219f216d62aed49ba837fda74
+oid sha256:df5490f404d7503d1542b2a546c472e06d9a5815189ec201cfc42790d6b07a3c
 size 6843306