{"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.11.13"},"kaggle":{"accelerator":"nvidiaTeslaT4","dataSources":[{"sourceId":108644,"databundleVersionId":13168647,"sourceType":"competition"}],"dockerImageVersionId":31090,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true},"papermill":{"default_parameters":{},"duration":124.346428,"end_time":"2025-07-29T20:44:37.054638","environment_variables":{},"exception":null,"input_path":"__notebook__.ipynb","output_path":"__notebook__.ipynb","parameters":{},"start_time":"2025-07-29T20:42:32.708210","version":"2.6.0"},"widgets":{"application/vnd.jupyter.widget-state+json":{"state":{"2e166a8435ea4f42b1bbd4d2c40a6bd1":{"model_module":"@jupyter-widgets/controls","model_module_version":"2.0.0","model_name":"ProgressStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"2.0.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"2.0.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"325e21a69d0e46319304aef26749b1ae":{"model_module":"@jupyter-widgets/base","model_module_version":"2.0.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"2.0.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"2.0.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border_bottom":null,"border_left":null,"border_right":null,"border_top":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"6875a7d48c704f39a75d20c8c553d566":{"model_module":"@jupyter-widgets/controls","model_module_version":"2.0.0","model_name":"HTMLStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"2.0.0","_model_name":"HTMLStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"2.0.0","_view_name":"StyleView","background":null,"description_width":"","font_size":null,"text_color":null}},"68f8066f714d4114a4f46e2bc4830034":{"model_module":"@jupyter-widgets/controls","model_module_version":"2.0.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"2.0.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"2.0.0","_view_name":"HTMLView","description":"","description_allow_html":false,"layout":"IPY_MODEL_325e21a69d0e46319304aef26749b1ae","placeholder":"​","style":"IPY_MODEL_8909de22f1ff4a9a9043b07734dae04b","tabbable":null,"tooltip":null,"value":"🧠 Qwen2.5:14b Analysis: 100%"}},"70e50e6843c544d0b879a4fbc455ae6e":{"model_module":"@jupyter-widgets/controls","model_module_version":"2.0.0","model_name":"FloatProgressModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"2.0.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"2.0.0","_view_name":"ProgressView","bar_style":"success","description":"","description_allow_html":false,"layout":"IPY_MODEL_a68c717a3fcc42fdbe9174e055ccce18","max":120,"min":0,"orientation":"horizontal","style":"IPY_MODEL_2e166a8435ea4f42b1bbd4d2c40a6bd1","tabbable":null,"tooltip":null,"value":120}},"8909de22f1ff4a9a9043b07734dae04b":{"model_module":"@jupyter-widgets/controls","model_module_version":"2.0.0","model_name":"HTMLStyleModel","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"2.0.0","_model_name":"HTMLStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"2.0.0","_view_name":"StyleView","background":null,"description_width":"","font_size":null,"text_color":null}},"a68c717a3fcc42fdbe9174e055ccce18":{"model_module":"@jupyter-widgets/base","model_module_version":"2.0.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"2.0.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"2.0.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border_bottom":null,"border_left":null,"border_right":null,"border_top":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"bfc3001ca4c14208b2c9dc9681745fa8":{"model_module":"@jupyter-widgets/base","model_module_version":"2.0.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"2.0.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"2.0.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border_bottom":null,"border_left":null,"border_right":null,"border_top":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"da835dc7accb42b1b3cd8e706c492ef9":{"model_module":"@jupyter-widgets/controls","model_module_version":"2.0.0","model_name":"HBoxModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"2.0.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"2.0.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_68f8066f714d4114a4f46e2bc4830034","IPY_MODEL_70e50e6843c544d0b879a4fbc455ae6e","IPY_MODEL_e1ee56c707e54f258b7bf1c907fc040c"],"layout":"IPY_MODEL_fb825f4c4cce485688541ca8c0731cb8","tabbable":null,"tooltip":null}},"e1ee56c707e54f258b7bf1c907fc040c":{"model_module":"@jupyter-widgets/controls","model_module_version":"2.0.0","model_name":"HTMLModel","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"2.0.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"2.0.0","_view_name":"HTMLView","description":"","description_allow_html":false,"layout":"IPY_MODEL_bfc3001ca4c14208b2c9dc9681745fa8","placeholder":"​","style":"IPY_MODEL_6875a7d48c704f39a75d20c8c553d566","tabbable":null,"tooltip":null,"value":" 120/120 [00:00<00:00, 386.97it/s]"}},"fb825f4c4cce485688541ca8c0731cb8":{"model_module":"@jupyter-widgets/base","model_module_version":"2.0.0","model_name":"LayoutModel","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"2.0.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"2.0.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border_bottom":null,"border_left":null,"border_right":null,"border_top":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}}},"version_major":2,"version_minor":0}}},"nbformat_minor":4,"nbformat":4,"cells":[{"id":"dfd34547","cell_type":"code","source":"# Competition Setup: Banglish Sentiment Challenge\n# This environment comes with analytics libraries for the sentiment analysis task\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O\n\n# Competition data files are available in the \"../input/\" directory\n# Let's explore the dataset structure for the Banglish Sentiment Challenge\n\nimport os\nprint(\"=== Competition Dataset Files ===\")\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        filepath = os.path.join(dirname, filename)\n        print(f\"📁 {filepath}\")\n        \n        # Show file size for better understanding\n        if os.path.exists(filepath):\n            size = os.path.getsize(filepath)\n            print(f\"   Size: {size:,} bytes\")\n\nprint(\"\\n=== Expected Files ===\")\nprint(\"📊 example.csv - Sample dataset with labeled examples\")\nprint(\"🧪 test.csv - Test set for predictions (no labels)\")  \nprint(\"📤 submission.csv - Submission template\")\n\n# Note: This is a zero-shot learning challenge - no training data provided!","metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","execution":{"iopub.execute_input":"2025-07-30T05:11:15.811778Z","iopub.status.busy":"2025-07-30T05:11:15.811532Z","iopub.status.idle":"2025-07-30T05:11:17.139836Z","shell.execute_reply":"2025-07-30T05:11:17.139183Z","shell.execute_reply.started":"2025-07-30T05:11:15.811755Z"},"papermill":{"duration":1.532939,"end_time":"2025-07-29T20:42:38.238628","exception":false,"start_time":"2025-07-29T20:42:36.705689","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"a77f0c7e","cell_type":"code","source":"# =============================================================================\n# SECTION 1: OLLAMA SETUP FOR ZERO-SHOT SENTIMENT ANALYSIS\n# =============================================================================\n\nprint(\"🚀 Installing Ollama and required packages...\")\nprint(\"This will enable us to run Llama3 locally for sentiment classification\")\n\n!curl -fsSL https://ollama.com/install.sh | sh\n!pip install -qq pyngrok ollama\n\nprint(\"✅ Installation complete!\")","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:11:17.140924Z","iopub.status.busy":"2025-07-30T05:11:17.140552Z","iopub.status.idle":"2025-07-30T05:11:57.771754Z","shell.execute_reply":"2025-07-30T05:11:57.770828Z","shell.execute_reply.started":"2025-07-30T05:11:17.140896Z"},"papermill":{"duration":40.300349,"end_time":"2025-07-29T20:43:18.541962","exception":false,"start_time":"2025-07-29T20:42:38.241613","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"da25ab33","cell_type":"code","source":"import subprocess\nimport os\nimport time\n\ndef start_ollama_server_with_gpu():\n    \"\"\"Starts Ollama server optimized for 2x T4 GPU setup.\"\"\"\n    print(\"🔧 Starting Ollama server with GPU acceleration for Banglish sentiment analysis...\")\n    \n    # Set GPU environment variables for optimal performance\n    gpu_env = os.environ.copy()\n    gpu_env['CUDA_VISIBLE_DEVICES'] = '0,1'  # Use both T4 GPUs\n    gpu_env['OLLAMA_NUM_PARALLEL'] = '2'     # Parallel processing on both GPUs\n    gpu_env['OLLAMA_MAX_LOADED_MODELS'] = '1'  # Keep one large model loaded\n    \n    print(\"🎯 GPU Configuration:\")\n    print(\"   • CUDA_VISIBLE_DEVICES: 0,1 (Both T4 GPUs)\")\n    print(\"   • OLLAMA_NUM_PARALLEL: 2 (Parallel processing)\")\n    print(\"   • Optimized for 14B parameter models\")\n    \n    # Ollama installation path\n    ollama_path = '/usr/local/bin/ollama'\n    if not os.path.exists(ollama_path):\n        print(\"⚠️  Ollama not found at /usr/local/bin, trying system PATH...\")\n        ollama_path = 'ollama'\n    \n    try:\n        # Check if server is already running\n        try:\n            subprocess.run(['pgrep', '-f', 'ollama serve'], check=True, \n                          stdout=subprocess.PIPE, stderr=subprocess.PIPE)\n            print(\"🔄 Ollama server already running, checking GPU status...\")\n        except subprocess.CalledProcessError:\n            print(\"🚀 Starting Ollama server with GPU optimization...\")\n            # Start with GPU environment\n            subprocess.Popen([ollama_path, 'serve'], env=gpu_env)\n            print(\"✅ Ollama server started with GPU acceleration!\")\n        \n        print(\"🕐 Initializing GPU memory and model cache...\")\n        \n    except Exception as e:\n        print(f\"❌ Error starting Ollama server: {e}\")\n        return False\n    \n    return True\n\ndef verify_gpu_setup():\n    \"\"\"Verify GPU detection and availability.\"\"\"\n    print(\"\\n🔍 GPU Verification:\")\n    try:\n        # Check NVIDIA GPUs\n        result = subprocess.run(['nvidia-smi', '--query-gpu=index,name,memory.total,memory.free', \n                               '--format=csv,noheader,nounits'], \n                               capture_output=True, text=True)\n        \n        if result.returncode == 0:\n            gpu_info = result.stdout.strip().split('\\n')\n            print(\"📊 Detected GPUs:\")\n            for i, gpu in enumerate(gpu_info):\n                print(f\"   GPU {i}: {gpu}\")\n            \n            if len(gpu_info) >= 2:\n                print(\"✅ Dual GPU setup confirmed for optimal large model performance\")\n                return True\n            else:\n                print(\"⚠️  Less than 2 GPUs detected\")\n        else:\n            print(\"❌ No NVIDIA GPUs detected\")\n            \n    except Exception as e:\n        print(f\"❌ GPU verification failed: {e}\")\n    \n    return False\n\n# Start optimized server\nif start_ollama_server_with_gpu():\n    # Verify GPU setup\n    gpu_ok = verify_gpu_setup()\n    \n    # Extended initialization time for large models\n    initialization_time = 12 if gpu_ok else 8\n    print(f\"⏱️  Waiting {initialization_time}s for GPU initialization...\")\n    time.sleep(initialization_time)\n    \n    if gpu_ok:\n        print(\"🎯 Server ready for GPU-accelerated sentiment analysis!\")\n        print(\"🚀 Optimized for Qwen2.5:14b and other large models\")\n    else:\n        print(\"⚠️  Server started but GPU optimization may be limited\")\nelse:\n    print(\"💥 Failed to start server - check installation and GPU drivers\")","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:11:57.774012Z","iopub.status.busy":"2025-07-30T05:11:57.773768Z","iopub.status.idle":"2025-07-30T05:12:09.821186Z","shell.execute_reply":"2025-07-30T05:12:09.820501Z","shell.execute_reply.started":"2025-07-30T05:11:57.773990Z"},"papermill":{"duration":12.069261,"end_time":"2025-07-29T20:43:30.625168","exception":false,"start_time":"2025-07-29T20:43:18.555907","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"292da788","cell_type":"code","source":"# =============================================================================\n# SECTION 2: GPU DETECTION & ADVANCED MODEL SETUP FOR MULTILINGUAL SENTIMENT\n# =============================================================================\n\nprint(\"🔥 Checking GPU availability for enhanced performance...\")\n\n# Check GPU detection\n!nvidia-smi\n\nprint(\"\\n📊 GPU Memory and Configuration:\")\n!nvidia-smi --query-gpu=name,memory.total,memory.free --format=csv\n\nprint(\"\\n🚀 Setting up advanced multilingual model for Banglish sentiment analysis...\")\nprint(\"Using Qwen2.5:14b - Excellent model for multilingual tasks including Bangla\")\n\n# Pull Qwen2.5:14b - Proven model with superior Bangla understanding\nprint(\"📥 Downloading Qwen2.5:14b model (this may take 10-15 minutes)...\")\n!ollama pull qwen2.5:14b\n\nprint(\"\\n✅ Qwen2.5:14b model ready!\")\nprint(\"🌍 This model excels at:\")\nprint(\"   • Excellent Bangla language understanding\")\nprint(\"   • Superior English language processing\") \nprint(\"   • Enhanced code-switching between languages\")\nprint(\"   • Reliable sentiment analysis tasks\")\nprint(\"   • Strong multilingual context comprehension\")\nprint(\"   • Good emoji and emoticon interpretation\")\n\nprint(\"\\n🔧 Verifying GPU utilization for optimal performance...\")\nprint(\"Expected: 2x T4 GPUs should be detected and utilized by Ollama\")\n\n# Alternative models to consider if Qwen2.5:14b is not available:\nprint(\"\\n📋 Model Information:\")\nprint(\"Primary: qwen2.5:14b (14B parameters) - Proven excellent for Banglish\")\nprint(\"Fallback: llama3.1:8b (8B parameters) - Good multilingual support\")\nprint(\"GPU Optimized: Uses both T4 GPUs for faster inference\")","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:12:09.824646Z","iopub.status.busy":"2025-07-30T05:12:09.823975Z","iopub.status.idle":"2025-07-30T05:13:10.625406Z","shell.execute_reply":"2025-07-30T05:13:10.624656Z","shell.execute_reply.started":"2025-07-30T05:12:09.824620Z"},"papermill":{"duration":61.520191,"end_time":"2025-07-29T20:44:32.160006","exception":false,"start_time":"2025-07-29T20:43:30.639815","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"ee914bd0","cell_type":"code","source":"# =============================================================================\n# GPU & MODEL VERIFICATION FOR OPTIMAL PERFORMANCE\n# =============================================================================\n\nprint(\"🔍 Verifying GPU detection and model availability:\")\nprint(\"=\" * 60)\n\n# Check Ollama models\nprint(\"📋 Available Ollama Models:\")\n!ollama list\n\nprint(\"\\n🎯 GPU Detection by Ollama:\")\n# Check if Ollama detects GPUs properly\nimport subprocess\ntry:\n    result = subprocess.run(['nvidia-smi', '-L'], capture_output=True, text=True)\n    gpu_info = result.stdout\n    print(\"🖥️ GPU Detection:\")\n    print(gpu_info)\n    \n    gpu_count = gpu_info.count('GPU')\n    print(f\"📊 Total GPUs detected: {gpu_count}\")\n    \n    if gpu_count >= 2:\n        print(\"✅ Dual GPU setup confirmed - Optimal for large model inference\")\n    else:\n        print(\"⚠️  Expected 2 GPUs, but found fewer\")\n        \nexcept Exception as e:\n    print(f\"❌ Error checking GPU: {e}\")\n\nprint(\"\\n🔥 Model Readiness Check:\")\nprint(\"✅ Qwen3:14b should be listed above for optimal Banglish sentiment analysis\")\nprint(\"🎯 This latest model provides superior performance for:\")\nprint(\"   • Advanced Bengali script recognition\")\nprint(\"   • Enhanced code-switching detection\")\nprint(\"   • Superior contextual sentiment understanding\")\nprint(\"   • Advanced emoji-aware analysis\")","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:14:51.931535Z","iopub.status.busy":"2025-07-30T05:14:51.930857Z","iopub.status.idle":"2025-07-30T05:14:52.105310Z","shell.execute_reply":"2025-07-30T05:14:52.104483Z","shell.execute_reply.started":"2025-07-30T05:14:51.931511Z"},"papermill":{"duration":0.222164,"end_time":"2025-07-29T20:44:32.424979","exception":false,"start_time":"2025-07-29T20:44:32.202815","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"59fff298","cell_type":"code","source":"# =============================================================================\n# SECTION 3: COMPETITION DATA EXPLORATION  \n# =============================================================================\n\nprint(\"📊 Loading and exploring competition datasets...\")\n\n# Load the datasets\ntry:\n    # Load example dataset (with labels for understanding format)\n    if os.path.exists('/kaggle/input/binary-biplob-can-you-decode-emotions/bangla/example.csv'):\n        example_df = pd.read_csv('/kaggle/input/binary-biplob-can-you-decode-emotions/bangla/example.csv')\n        print(\"✅ Example dataset loaded successfully!\")\n        print(f\"📈 Example dataset shape: {example_df.shape}\")\n        print(\"\\n🔍 Example data preview:\")\n        print(example_df.head())\n        if 'label' in example_df.columns:\n            print(f\"\\n📊 Label distribution in examples:\")\n            print(example_df['label'].value_counts())\n    \n    # Load test dataset (what we need to predict)\n    if os.path.exists('/kaggle/input/binary-biplob-can-you-decode-emotions/bangla/test.csv'):\n        test_df = pd.read_csv('/kaggle/input/binary-biplob-can-you-decode-emotions/bangla/test.csv')\n        print(f\"\\n✅ Test dataset loaded successfully!\")\n        print(f\"� Test dataset shape: {test_df.shape}\")\n        print(\"\\n🔍 Test data preview:\")\n        print(test_df.head())\n        \n        # Analyze text characteristics\n        print(f\"\\n📏 Text length statistics:\")\n        test_df['text_length'] = test_df['text'].str.len()\n        print(test_df['text_length'].describe())\n        \n    else:\n        print(\"⚠️  Test dataset not found - using sample data\")\n        test_df = pd.DataFrame({\n            'id': ['sample_1', 'sample_2', 'sample_3'], \n            'text': ['আজকে weather nice 😊', 'feeling sad today 😢', 'kaj cholche normally']\n        })\n\nexcept Exception as e:\n    print(f\"❌ Error loading datasets: {e}\")\n    print(\"📝 Creating sample data for testing...\")\n    test_df = pd.DataFrame({\n        'id': ['sample_1', 'sample_2', 'sample_3'], \n        'text': ['আজকে weather nice 😊', 'feeling sad today 😢', 'kaj cholche normally']\n    })\n\nprint(f\"\\n� Ready to process {len(test_df)} samples for sentiment prediction!\")","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:14:59.350271Z","iopub.status.busy":"2025-07-30T05:14:59.349615Z","iopub.status.idle":"2025-07-30T05:14:59.391826Z","shell.execute_reply":"2025-07-30T05:14:59.391166Z","shell.execute_reply.started":"2025-07-30T05:14:59.350233Z"},"papermill":{"duration":0.102176,"end_time":"2025-07-29T20:44:32.620555","exception":false,"start_time":"2025-07-29T20:44:32.518379","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"bd0af602","cell_type":"code","source":"# =============================================================================\n# DATASET CONFIGURATION\n# =============================================================================\n\nprint(\"🎯 BANGLISH SENTIMENT CHALLENGE - DATASET CONFIGURATION\")\nprint(\"=\" * 60)\n\n# Using the specified test dataset path\ntest_dataset_path = '/kaggle/input/binary-biplob-can-you-decode-emotions/bangla/test.csv'\nprint(f\"📊 Test Dataset Path: {test_dataset_path}\")\n\n# Verify dataset accessibility\nimport os\nif os.path.exists(test_dataset_path):\n    print(\"✅ Test dataset found and accessible\")\n    try:\n        # Quick preview without loading full dataset\n        import pandas as pd\n        sample_df = pd.read_csv(test_dataset_path, nrows=3)\n        print(f\"📋 Dataset columns: {list(sample_df.columns)}\")\n        print(f\"🔍 Sample entries:\")\n        for _, row in sample_df.iterrows():\n            print(f\"   ID: {row['id']}, Text preview: {str(row['text'])[:50]}...\")\n    except Exception as e:\n        print(f\"⚠️  Preview error: {e}\")\nelse:\n    print(\"⚠️  Test dataset not found at specified path\")\n    print(\"🔧 Ensure the dataset is available at the Kaggle input location\")\n\nprint(f\"\\n🎯 Model Configuration:\")\nprint(\"• Primary Model: qwen3:14b (Latest and best for Banglish)\")\nprint(\"• Fallback Model: llama3.1:8b\")\nprint(\"• Output Format: id,label (competition standard)\")\nprint(\"• Labels: positive, negative, neutral\")\n\nprint(f\"\\n🚀 Ready to process the full dataset with advanced sentiment analysis!\")","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:15:12.328910Z","iopub.status.busy":"2025-07-30T05:15:12.328234Z","iopub.status.idle":"2025-07-30T05:15:12.340526Z","shell.execute_reply":"2025-07-30T05:15:12.339858Z","shell.execute_reply.started":"2025-07-30T05:15:12.328889Z"},"trusted":true},"outputs":[],"execution_count":null},{"id":"8c794685","cell_type":"code","source":"# =============================================================================\n# SECTION 4: ADVANCED BANGLISH SENTIMENT CLASSIFICATION WITH QWEN2.5:14B\n# =============================================================================\n\nimport pandas as pd\nimport ollama\nimport re\nfrom tqdm.auto import tqdm\n\nprint(\"🎯 BANGLISH SENTIMENT CHALLENGE - QWEN2.5:14B SIMPLE SOLUTION\")\nprint(\"=\" * 60)\n\ndef get_banglish_sentiment(text_input, client):\n    \n    # Simple and direct prompt - what worked for 0.75 accuracy\n    prompt = f\"\"\"Analyze sentiment: positive, negative, or neutral\n\nText: {text_input}\n\nSentiment:\"\"\"\n\n    try:\n        response = client.generate(\n            model='qwen2.5:14b',  # Using qwen2.5:14b as you found it best\n            prompt=prompt,\n            options={\n                'temperature': 0.1,       # Low but not too restrictive\n                'num_predict': 5,         # Allow a bit more output\n                'top_p': 0.9,            # Higher for natural language\n                'repeat_penalty': 1.1,   # Minimal repetition penalty\n                'num_ctx': 2048,         # Smaller context for efficiency\n                'num_gpu': 2             # Utilize both T4 GPUs\n            }\n        )\n        \n        sentiment = response['response'].strip().lower()\n        \n        # Simple direct parsing - no complex regex needed\n        if 'positive' in sentiment:\n            return \"positive\"\n        elif 'negative' in sentiment:\n            return \"negative\"\n        elif 'neutral' in sentiment:\n            return \"neutral\"\n        else:\n            # If unclear, try first word\n            first_word = sentiment.split()[0] if sentiment.split() else \"\"\n            if first_word.startswith('pos'):\n                return \"positive\"\n            elif first_word.startswith('neg'):\n                return \"negative\"\n            else:\n                return \"neutral\"  # Default to neutral for unclear cases\n                \n    except Exception as e:\n        print(f\"⚠️  Error with Qwen2.5:14b: {e}\")\n        # Simple fallback - no complex alternative models\n        return \"neutral\"\n\n# Initialize Ollama client with GPU optimization\ntry:\n    client = ollama.Client(host='http://127.0.0.1:11434')\n    models = client.list()\n    print(\"✅ Connected to Ollama server with GPU acceleration\")\n    \n    # Verify our target model is available\n    try:\n        # Handle different response formats from Ollama API\n        if hasattr(models, 'models'):\n            model_list = models.models\n        elif isinstance(models, dict) and 'models' in models:\n            model_list = models['models']\n        else:\n            model_list = models\n        \n        # Extract model names safely\n        model_names = []\n        for model in model_list:\n            if hasattr(model, 'name'):\n                model_names.append(model.name)\n            elif isinstance(model, dict) and 'name' in model:\n                model_names.append(model['name'])\n            elif hasattr(model, 'model'):\n                model_names.append(model.model)\n            elif isinstance(model, dict) and 'model' in model:\n                model_names.append(model['model'])\n        \n        print(f\"🔍 Available models: {model_names}\")\n        \n        if 'qwen2.5:14b' in model_names:\n            print(\"🎯 Qwen2.5:14b model confirmed - Optimal for Banglish analysis\")\n        else:\n            print(\"⚠️  Qwen2.5:14b not found, will attempt to use available models\")\n            \n    except Exception as model_error:\n        print(f\"⚠️  Model verification error: {model_error}\")\n        print(\"🔄 Proceeding with available models...\")\n    \nexcept Exception as e:\n    print(f\"❌ Cannot connect to Ollama: {e}\")\n    print(\"🔧 Make sure Ollama server is running with GPU support!\")\n    raise\n\n# Load test data from specified path\ntry:\n    test_df = pd.read_csv('/kaggle/input/binary-biplob-can-you-decode-emotions/bangla/test.csv')\n    print(f\"📊 Loaded test data: {test_df.shape[0]} samples\")\n    print(f\"📋 Columns: {list(test_df.columns)}\")\n    print(f\"🔍 Sample data preview:\")\n    print(test_df.head())\n    \nexcept FileNotFoundError:\n    print(f\"❌ Test dataset not found at: /kaggle/input/binary-biplob-can-you-decode-emotions/bangla/test.csv\")\n    print(\"🔧 Please ensure the file exists at the specified path\")\n    raise\nexcept Exception as e:\n    print(f\"❌ Error loading test dataset: {e}\")\n    raise\n\nprint(f\"\\n🚀 Starting Qwen2.5:14b simple sentiment analysis...\")\nprint(f\"⚡ Processing {len(test_df)} samples with direct prompting approach...\")\n\n# Apply simple sentiment analysis\ntqdm.pandas(desc=\"🧠 Qwen2.5:14b Simple Analysis\")\ntest_df['predicted_label'] = test_df['text'].progress_apply(\n    lambda x: get_banglish_sentiment(x, client)\n)\n\n# Results summary\nprint(f\"\\n📊 SIMPLE SENTIMENT ANALYSIS COMPLETE!\")\nprint(\"=\" * 50)\nprint(\"📈 Prediction distribution:\")\nprint(test_df['predicted_label'].value_counts())\n\nprint(f\"\\n🔍 Sample predictions with simple Qwen2.5:14b:\")\nfor _, row in test_df.head(10).iterrows():\n    text_preview = row['text'][:70] + \"...\" if len(row['text']) > 70 else row['text']\n    print(f\"📝 {row['predicted_label'].upper()}: {text_preview}\")\n\n# Prepare submission with exact format requested\nsubmission_df = test_df[['id', 'predicted_label']].copy()\nsubmission_df = submission_df.rename(columns={'predicted_label': 'label'})\nsubmission_df.to_csv('submission.csv', index=False)\n\nprint(f\"\\n✅ SUBMISSION.CSV GENERATED!\")\nprint(f\"💾 Format: id,label\")\nprint(f\"📊 Total predictions: {len(submission_df)}\")\nprint(f\"🎯 Using simple direct prompting with Qwen2.5:14b for better accuracy\")\n\n# Show sample of final submission format\nprint(f\"\\n📄 SUBMISSION.CSV PREVIEW:\")\nprint(\"id,label\")\nfor _, row in submission_df.head(5).iterrows():\n    print(f\"{row['id']},{row['label']}\")\n\nprint(\"🏆 Ready for competition submission!\")","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:16:12.545315Z","iopub.status.busy":"2025-07-30T05:16:12.544461Z","iopub.status.idle":"2025-07-30T05:30:18.721013Z","shell.execute_reply":"2025-07-30T05:30:18.720404Z","shell.execute_reply.started":"2025-07-30T05:16:12.545287Z"},"papermill":{"duration":1.541731,"end_time":"2025-07-29T20:44:34.204704","exception":false,"start_time":"2025-07-29T20:44:32.662973","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"76dd435d","cell_type":"code","source":"# =============================================================================\n# SECTION 5: SUBMISSION VALIDATION & FINAL CHECKS\n# =============================================================================\n\nprint(\"🔍 VALIDATING SUBMISSION FOR BANGLISH SENTIMENT CHALLENGE\")\nprint(\"=\" * 60)\n\n# Load and validate submission file\ntry:\n    submission_check = pd.read_csv('submission.csv')\n    \n    print(\"✅ Submission file loaded successfully!\")\n    print(f\"📊 Submission shape: {submission_check.shape}\")\n    print(f\"📋 Columns: {list(submission_check.columns)}\")\n    \n    # Validate required columns\n    required_cols = ['id', 'label']\n    missing_cols = [col for col in required_cols if col not in submission_check.columns]\n    \n    if missing_cols:\n        print(f\"❌ Missing required columns: {missing_cols}\")\n    else:\n        print(\"✅ All required columns present\")\n    \n    # Validate label values\n    valid_labels = {'positive', 'negative', 'neutral'}\n    unique_labels = set(submission_check['label'].unique())\n    invalid_labels = unique_labels - valid_labels\n    \n    if invalid_labels:\n        print(f\"⚠️  Invalid labels found: {invalid_labels}\")\n        print(\"🔧 Valid labels are: positive, negative, neutral\")\n    else:\n        print(\"✅ All labels are valid\")\n    \n    # Check for missing values\n    missing_count = submission_check.isnull().sum().sum()\n    if missing_count > 0:\n        print(f\"⚠️  Found {missing_count} missing values\")\n    else:\n        print(\"✅ No missing values\")\n    \n    # Summary statistics\n    print(f\"\\n📊 FINAL SUBMISSION SUMMARY:\")\n    print(f\"📝 Total predictions: {len(submission_check)}\")\n    print(f\"📈 Label distribution:\")\n    label_counts = submission_check['label'].value_counts()\n    for label, count in label_counts.items():\n        percentage = (count / len(submission_check)) * 100\n        print(f\"   {label}: {count} ({percentage:.1f}%)\")\n    \n    # Show sample submission format\n    print(f\"\\n📄 SAMPLE SUBMISSION FORMAT:\")\n    print(\"id,label\")\n    for _, row in submission_check.head(5).iterrows():\n        print(f\"{row['id']},{row['label']}\")\n    \n    # Calculate macro F1 readiness\n    print(f\"\\n🎯 COMPETITION READINESS:\")\n    print(\"✅ Format matches submission requirements (id,label)\")\n    print(\"✅ Uses macro-averaged F1-score evaluation\")\n    print(\"✅ Zero-shot approach with Qwen3:14b (latest model)\")\n    print(\"✅ Handles Banglish code-switching\")\n    \n    print(f\"\\n🏆 SUBMISSION READY FOR UPLOAD!\")\n    print(\"📤 File: submission.csv\")\n    print(\"🎪 Competition: Banglish Sentiment Challenge\")\n    \nexcept FileNotFoundError:\n    print(\"❌ Submission file not found!\")\n    print(\"🔧 Run the previous cells to generate predictions\")\n    \nexcept Exception as e:\n    print(f\"❌ Error validating submission: {e}\")\n\nprint(\"\\n\" + \"=\" * 60)\nprint(\"🎯 BANGLISH SENTIMENT CHALLENGE SOLUTION COMPLETE\")\nprint(\"=\" * 60)","metadata":{"execution":{"iopub.execute_input":"2025-07-30T05:30:18.722586Z","iopub.status.busy":"2025-07-30T05:30:18.722365Z","iopub.status.idle":"2025-07-30T05:30:18.737326Z","shell.execute_reply":"2025-07-30T05:30:18.736578Z","shell.execute_reply.started":"2025-07-30T05:30:18.722571Z"},"papermill":{"duration":0.058753,"end_time":"2025-07-29T20:44:34.308894","exception":false,"start_time":"2025-07-29T20:44:34.250141","status":"completed"},"tags":[],"trusted":true},"outputs":[],"execution_count":null},{"id":"d90fee54","cell_type":"code","source":"","metadata":{"papermill":{"duration":0.043091,"end_time":"2025-07-29T20:44:34.395912","exception":false,"start_time":"2025-07-29T20:44:34.352821","status":"completed"},"tags":[]},"outputs":[],"execution_count":null}]}