End of training

README.md

@@ -22,7 +22,6 @@ adapter: lora
 base_model: sethuiyer/Medichat-Llama3-8B
 bf16: false
 chat_template: llama3
-dataset_prepared_path: null
 datasets:
 - data_files:
   - 95cbf511558bbb4b_train_data.json
@@ -42,8 +41,8 @@ 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
@@ -66,7 +65,7 @@ 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/95cbf511558bbb4b_train_data.json
 model_type: AutoModelForCausalLM
@@ -77,7 +76,7 @@ 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
@@ -91,7 +90,7 @@ wandb_name: b9ccccb9-64a3-4207-a2ed-fb5da2aeefd2
 wandb_project: Gradients-On-Demand
 wandb_run: your_name
 wandb_runid: b9ccccb9-64a3-4207-a2ed-fb5da2aeefd2
-warmup_steps: 10
+warmup_steps: 0
 weight_decay: 0.0
 xformers_attention: null
 
@@ -103,7 +102,7 @@ xformers_attention: null
 
 This model is a fine-tuned version of [sethuiyer/Medichat-Llama3-8B](https://huggingface.co/sethuiyer/Medichat-Llama3-8B) on the None dataset.
 It achieves the following results on the evaluation set:
-- Loss: 2.1259
+- Loss: 1.0635
 
 ## Model description
 
@@ -130,18 +129,14 @@ 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: 333
 - mixed_precision_training: Native AMP
 
 ### Training results
 
-| Training Loss | Epoch  | Step | Validation Loss |
-|:-------------:|:------:|:----:|:---------------:|
-| 2.0249        | 0.0030 | 1    | 2.2815          |
-| 2.3406        | 0.0090 | 3    | 2.2815          |
-| 2.6119        | 0.0180 | 6    | 2.2540          |
-| 1.7983        | 0.0270 | 9    | 2.1259          |
+| Training Loss | Epoch | Step | Validation Loss |
+|:-------------:|:-----:|:----:|:---------------:|
+| 0.8482        | 1.0   | 333  | 1.0635          |
 
 
 ### Framework versions

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a94317c295e7b8221749385f9f9f7ad18bdcf24c212100fe9488e1bbc1cbb50e
+oid sha256:104c46145be25520cf67693749ef0de394110dde74006158300b60389d152d47
 size 84047370