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/079cecae2ab93995_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: 9b36f719-bc20-45e9-a7b0-55362538f00a
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 9b36f719-bc20-45e9-a7b0-55362538f00a
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # 87c402c9-1a84-6e93-85d0-dd33c92ea7e6
 
 This model is a fine-tuned version of [zake7749/gemma-2-2b-it-chinese-kyara-dpo](https://huggingface.co/zake7749/gemma-2-2b-it-chinese-kyara-dpo) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.6373
 
 ## 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        | 0.0004 | 1    | 3.5400          |
+| 2.7103        | 0.0037 | 10   | 2.0698          |
+| 0.9326        | 0.0074 | 20   | 0.7796          |
+| 0.7495        | 0.0110 | 30   | 0.6703          |
+| 0.6783        | 0.0147 | 40   | 0.6427          |
+| 0.6324        | 0.0184 | 50   | 0.6373          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:98c9ba47f26e793d810d37a4de1462acb2e701de94e9e607b2bdb74cf466eac5
+oid sha256:68c1ec32c3723f5a30cfb8143f5dcb3856aa697b08b5200459ccbfcd2ce704d2
 size 41664442