Update README.md

README.md

@@ -59,46 +59,39 @@ Peft Config :
 
 {
  'Technqiue' : 'QLORA',
- 
  'rank': 256,
- 
  'target_modules' : ["q_proj", "k_proj", "v_proj", "o_proj","gate_proj", "up_proj", "down_proj",],
- 
  'lora_alpha' : 128,
- 
  'lora_dropout' : 0, 
- 
  'bias': "none",    
-
 }
     
 Hyperparameters:
 
 {
     "epochs": 3,
-    
     "evaluation_strategy": "epoch",
-    
     "gradient_checkpointing": True,
-    
     "max_grad_norm" : 0.3,
-    
     "optimizer" : "adamw_torch_fused",
-    
     "learning_rate" : 2e-4,
-    
     "lr_scheduler_type": "constant",
-    
     "warmup_ratio" : 0.03,
-    
-    "per_device_train_batch_size" : 8,  
-    
-    "per_device_eval_batch_size" : 8,
-    
+    "per_device_train_batch_size" : 4,  
+    "per_device_eval_batch_size" : 4,
     "gradient_accumulation_steps" : 4
-
 }
 ```
+Model was trained on 1xA100 80GB, below loss and memory consmuption details:
+{'eval_loss': 0.9614351987838745, 'eval_runtime': 244.0411, 'eval_samples_per_second': 2.663, 'eval_steps_per_second': 0.668, 'epoch': 3.0}
+{'train_runtime': 19718.5285, 'train_samples_per_second': 0.781, 'train_steps_per_second': 0.049, 'train_loss': 0.8241131883172602, 'epoch': 3.0}
+Total training time 19720.924563884735
+328.64 minutes used for training.
+Peak reserved memory = 35.789 GB.
+Peak reserved memory for training = 27.848 GB.
+Peak reserved memory % of max memory = 45.216 %.
+Peak reserved memory for training % of max memory = 35.183 %.
+
 
 ## Evaluation