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/886b76d2d77be502_train_data.json
 model_type: AutoModelForCausalLM
@@ -91,7 +91,7 @@ wandb_name: 7930e7df-e1ad-4e4d-a081-7b1a8312e5ba
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 7930e7df-e1ad-4e4d-a081-7b1a8312e5ba
-warmup_steps: 1
+warmup_steps: 10
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,6 +102,8 @@ xformers_attention: null
 # 1b27b1c8-4fc0-4f6b-8a83-bbf4b725544e
 
 This model is a fine-tuned version of [NousResearch/Nous-Hermes-llama-2-7b](https://huggingface.co/NousResearch/Nous-Hermes-llama-2-7b) on the None dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.0004
 
 ## 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: 30
 
 ### Training results
 
 | Training Loss | Epoch  | Step | Validation Loss |
 |:-------------:|:------:|:----:|:---------------:|
 | No log        | 0.0336 | 1    | 0.4334          |
+| 0.4071        | 0.2017 | 6    | 0.3601          |
+| 0.2676        | 0.4034 | 12   | 0.0196          |
+| 0.0289        | 0.6050 | 18   | 0.0011          |
+| 0.0018        | 0.8067 | 24   | 0.0005          |
+| 0.0005        | 1.0168 | 30   | 0.0004          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5da20deee5f16df3d0b1cffb1bba1f8bacffbb867d67c4bf7bebe0db9061ba2a
+oid sha256:af0154ea1adff37efea2a9d2f94a75e5f6f298221eb78c4e9ecaeb6852c2effd
 size 604420306