End of training

README.md

@@ -22,7 +22,6 @@ adapter: lora
 base_model: unsloth/Llama-3.2-1B-Instruct
 bf16: false
 chat_template: llama3
-dataset_prepared_path: null
 datasets:
 - data_files:
   - 51bfccbb1c68b256_train_data.json
@@ -41,8 +40,8 @@ deepspeed: null
 early_stopping_patience: null
 eval_max_new_tokens: 128
 eval_table_size: null
-evals_per_epoch: 4
-flash_attention: true
+evals_per_epoch: 1
+flash_attention: false
 fp16: true
 fsdp: null
 fsdp_config: null
@@ -65,7 +64,7 @@ lora_model_dir: null
 lora_r: 8
 lora_target_linear: true
 lr_scheduler: cosine
-max_steps: 10
+max_steps: 1000
 micro_batch_size: 1
 mlflow_experiment_name: /tmp/51bfccbb1c68b256_train_data.json
 model_type: AutoModelForCausalLM
@@ -76,7 +75,7 @@ pad_to_sequence_len: true
 resume_from_checkpoint: null
 s2_attention: null
 sample_packing: false
-saves_per_epoch: 4
+saves_per_epoch: 1
 sequence_len: 1024
 strict: false
 tf32: false
@@ -90,7 +89,7 @@ wandb_name: 4d407371-e885-4538-9aa3-e0817fccb84f
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: 4d407371-e885-4538-9aa3-e0817fccb84f
-warmup_steps: 10
+warmup_steps: 0
 weight_decay: 0.0
 xformers_attention: null
 
@@ -102,7 +101,7 @@ xformers_attention: null
 
 This model is a fine-tuned version of [unsloth/Llama-3.2-1B-Instruct](https://huggingface.co/unsloth/Llama-3.2-1B-Instruct) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.9398
+- Loss: 0.8004
 
 ## Model description
 
@@ -129,18 +128,14 @@ The following hyperparameters were used during training:
 - total_train_batch_size: 4
 - optimizer: Use OptimizerNames.ADAMW_HF 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: 10
-- training_steps: 10
+- training_steps: 719
 - mixed_precision_training: Native AMP
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss |
-|:-------------:|:------:|:----:|:---------------:|
-| 1.3831        | 0.0014 | 1    | 1.0150          |
-| 0.8161        | 0.0042 | 3    | 1.0096          |
-| 0.8737        | 0.0083 | 6    | 0.9795          |
-| 0.5585        | 0.0125 | 9    | 0.9398          |
+| Training Loss | Epoch | Step | Validation Loss |
+|:-------------:|:-----:|:----:|:---------------:|
+| 1.321         | 1.0   | 719  | 0.8004          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4cd3152f857d7036d180982c9b7705ea4d2a1029b8997071ce41cc3877da2701
+oid sha256:7edb84550813825630cac200c53d182424fce5c2f31c70a2298a4a5798681278
 size 22624970