from datasets import load_dataset
import os
# List of datasets to download
datasets_to_download = [
    'arc_challenge',           # arc-challenge-indic
    'audio_evals',             # audiollm-evals  
    'boolq_indic',             # boolq-indic
    'contextual_asr_benchmark', # contextual_asr_benchmark
    'gsm8k_indic',             # gsm8k-indic
    'mmlu_indic',              # mmlu-indic
    'olmOCR_Bench_English',    # olmOCR-Bench-English
    'samvaad_hi_v1',           # samvaad-hi-v1
    'sarvam_dub_benchmark_set', # sarvam-dub-benchmark-set
    'trivia_qa_indic_mcq',     # trivia-qa-indic-mcq
    'tts_general_benchmark',   # tts-general-benchmark
    'tts_robustness_benchmark'  # tts-robustness-benchmark
]
# Create directory for datasets
os.makedirs('datasets', exist_ok=True)
# Download each dataset
for dataset_name in datasets_to_download:
    try:
        print(f"Downloading {dataset_name}...")
        dataset = load_dataset(dataset_name)
        
        # Save to disk
        output_path = f'datasets/{dataset_name}'
        if isinstance(dataset, dict):
            # If it's a dictionary (train/test/validation splits)
            for split_name, split_data in dataset.items():
                split_path = f'{output_path}_{split_name}'
                split_data.save_to_disk(split_path)
                print(f"  Saved {split_name} split to {split_path}")
        else:
            # If it's a single dataset
            dataset.save_to_disk(output_path)
            print(f"  Saved to {output_path}")
            
    except Exception as e:
        print(f"  Failed to download {dataset_name}: {str(e)}")
        print(f"  Trying alternative names...")
        
        # Try some common variations
        alternatives = [
            dataset_name.replace('_indic', ''),
            dataset_name.replace('_indic', '_indian'),
            dataset_name.replace('_benchmark', '_benchmarks'),
            dataset_name.replace('_general', '_general_bench'),
            dataset_name.replace('_robustness', '_robust'),
        ]
        
        for alt_name in alternatives:
            try:
                print(f"  Trying: {alt_name}")
                dataset = load_dataset(alt_name)
                output_path = f'datasets/{alt_name}'
                if isinstance(dataset, dict):
                    for split_name, split_data in dataset.items():
                        split_path = f'{output_path}_{split_name}'
                        split_data.save_to_disk(split_path)
                        print(f"    Saved {split_name} split to {split_path}")
                else:
                    dataset.save_to_disk(output_path)
                    print(f"    Saved to {output_path}")
                break
            except:
                continue
print("Dataset download complete!")