Update README.md

README.md

@@ -82,56 +82,35 @@ The original protocol specifications have been:
 - Standardized to a consistent text format
 - Converted to JSONL for easy loading
 
-## Loading the Dataset
-
-```python
-from datasets import load_dataset
-
-dataset = load_dataset("your-username/spore-protocols")
-
-# Access a protocol
-print(dataset['train'][0]['text'])
-```
-
 ## Training Example
 
 ```python
-from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, Trainer
-
-model = AutoModelForCausalLM.from_pretrained("your-model")
-tokenizer = AutoTokenizer.from_pretrained("your-model")
-
-def tokenize_function(examples):
-    return tokenizer(examples["text"], truncation=True, max_length=512)
+from datasets import load_dataset
 
-tokenized_dataset = dataset.map(tokenize_function, batched=True)
+dataset = load_dataset("dassarthak18/spore-protocols")
 
 training_args = TrainingArguments(
-    output_dir="./results",
-    num_train_epochs=3,
-    per_device_train_batch_size=4,
-    learning_rate=2e-5,
+    # your args
 )
 
-trainer = Trainer(
+trainer = SFTTrainer(
     model=model,
     args=training_args,
-    train_dataset=tokenized_dataset["train"],
+    train_dataset=dataset['train'],
+    dataset_text_field="text",
 )
-
-trainer.train()
 ```
 
 ## Dataset Statistics
 
-- **Total Protocols**: ~50 protocols
+- **Total Protocols**: 49 protocols
 - **Protocol Families**: Includes variations, fixes, and modified versions of classic protocols
 - **Average Length**: Varies from simple 3-message protocols to complex multi-party protocols
 - **Language**: Formal protocol specification language
 
 ## Limitations
 
-- The dataset is relatively small (~50 protocols)
+- The dataset is relatively small (49 protocols)
 - Protocols are represented in a specific formal notation
 - Does not include natural language descriptions or security proofs
 - Limited to protocols documented in SPORE as of the collection date