Update README.md

Added 2b-it training content

README.md

@@ -2,14 +2,14 @@
 datasets:
 - saidines12/telugu_news_dataset
 base_model:
-- google/gemma-2b
+- google/gemma-2b-it
 ---
 
 
 
 # Model Card for Gemma-2B Telugu News Headline Generator
 
-This model is a fine-tuned version of Google's Gemma-2B model, optimized for generating Telugu news headlines from article content. It has been trained using Supervised Fine-Tuning (SFT) on a Telugu news dataset.
+This model is a fine-tuned version of Google's Gemma-2B Instruction model, optimized for generating Telugu news headlines from article content. It has been trained using Supervised Fine-Tuning (SFT) on a Telugu news dataset.
 
 ## Model Details
 
@@ -23,7 +23,7 @@ This model is a fine-tuned version of Google's Gemma-2B model, optimized for gen
 
 ### Model Sources
 - **Repository:** Hugging Face Hub
-- **Base Model:** google/gemma-2b
+- **Base Model:** google/gemma-2b-it
 
 ## Uses
 
@@ -77,13 +77,13 @@ headline = tokenizer.decode(outputs[0], skip_special_tokens=True)
 
 #### Training Hyperparameters
 - **Training regime:** FP16 mixed precision
-- **Batch size:** 4 per device
+- **Batch size:** 6 per device
 - **Gradient accumulation steps:** 4
 - **Learning rate:** 2e-4
-- **Maximum steps:** 30,000
+- **Maximum steps:** 20,000
 - **Warmup steps:** 25
 - **Optimizer:** AdamW
-- **Evaluation strategy:** Every 30000 steps
+- **Evaluation strategy:** Every 20000 steps 
 
 #### Hardware Specifications
 - GPU training with gradient checkpointing
@@ -100,7 +100,7 @@ headline = tokenizer.decode(outputs[0], skip_special_tokens=True)
 ## Technical Specifications
 
 ### Model Architecture and Objective
-- Base architecture: Gemma-2B
+- Base architecture: Gemma-2
 - Training objective: Supervised fine-tuning for headline generation
 - Gradient checkpointing enabled for memory efficiency
 - Optimized data loading with pinned memory