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/a1efc63f46bfa4cb_train_data.json
 model_type: AutoModelForCausalLM
@@ -93,7 +93,7 @@ wandb_name: e0a58fb5-adb8-4c21-a2df-12740dec5800
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: e0a58fb5-adb8-4c21-a2df-12740dec5800
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -104,6 +104,8 @@ xformers_attention: null
 # bc1f35d9-6b7f-726a-21bd-a196d8a9b283
 
 This model is a fine-tuned version of [huggyllama/llama-7b](https://huggingface.co/huggyllama/llama-7b) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 3.1076
 
 ## 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    | 3.5858          |
+| 3.5335        | 0.0013 | 10   | 3.4641          |
+| 3.3127        | 0.0027 | 20   | 3.2420          |
+| 3.1838        | 0.0040 | 30   | 3.1548          |
+| 3.1005        | 0.0054 | 40   | 3.1158          |
+| 3.1126        | 0.0067 | 50   | 3.1076          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48579261511c49cf53fa566ff0878c06046c4077d06da2fa870372cbc5112da2
+oid sha256:24ee1103701a5648da49a0d702950524efd5ba6a81fd0475f14cdf4ff4195c13
 size 80115210