| {"repo_id":"bigram_language_model","entity_id":"py:setup","uri":"program://bigram_language_model/module/setup#L1-L23","kind":"module","name":"setup","path":"setup.py","language":"python","start_line":1,"end_line":23,"context_start_line":1,"context_end_line":23,"code":"from setuptools import setup, find_packages\n\nsetup(\n name=\"bigram-lm\",\n version=\"0.1.0\",\n packages=find_packages(),\n install_requires=[\n \"torch>=2.0.0\",\n \"numpy>=1.21.0\",\n ],\n extras_require={\n 'test': [\n 'pytest>=7.0.0',\n 'pytest-cov>=4.0.0',\n ],\n },\n author=\"Your Name\",\n author_email=\"your.email@example.com\",\n description=\"A simple bigram language model implementation\",\n long_description=open(\"README.md\").read(),\n long_description_content_type=\"text/markdown\",\n python_requires=\">=3.7\",\n) ","source_hash":"bbd5fea0a69a2f9d71ead2e60fbb59c7d795c3a4bc1d6719275a61598e8205ef","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.conftest","uri":"program://bigram_language_model/module/tests.conftest#L1-L20","kind":"module","name":"tests.conftest","path":"tests/conftest.py","language":"python","start_line":1,"end_line":20,"context_start_line":1,"context_end_line":20,"code":"import pytest\nimport torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\n@pytest.fixture\ndef sample_text():\n return \"Hello, World!\"\n\n@pytest.fixture\ndef dataset(sample_text):\n return TextDataset(sample_text, sequence_length=3)\n\n@pytest.fixture\ndef model(dataset):\n return BigramLanguageModel(dataset.vocab_size)\n\n@pytest.fixture\ndef device():\n return 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"3b2c33bcd73f31217a9236b498c180e79716f55a1fead5651ce753267ac626cf","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.conftest.sample_text","uri":"program://bigram_language_model/function/tests.conftest.sample_text#L7-L8","kind":"function","name":"sample_text","path":"tests/conftest.py","language":"python","start_line":7,"end_line":8,"context_start_line":1,"context_end_line":20,"code":"import pytest\nimport torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\n@pytest.fixture\ndef sample_text():\n return \"Hello, World!\"\n\n@pytest.fixture\ndef dataset(sample_text):\n return TextDataset(sample_text, sequence_length=3)\n\n@pytest.fixture\ndef model(dataset):\n return BigramLanguageModel(dataset.vocab_size)\n\n@pytest.fixture\ndef device():\n return 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"3b2c33bcd73f31217a9236b498c180e79716f55a1fead5651ce753267ac626cf","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.conftest.dataset","uri":"program://bigram_language_model/function/tests.conftest.dataset#L11-L12","kind":"function","name":"dataset","path":"tests/conftest.py","language":"python","start_line":11,"end_line":12,"context_start_line":1,"context_end_line":20,"code":"import pytest\nimport torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\n@pytest.fixture\ndef sample_text():\n return \"Hello, World!\"\n\n@pytest.fixture\ndef dataset(sample_text):\n return TextDataset(sample_text, sequence_length=3)\n\n@pytest.fixture\ndef model(dataset):\n return BigramLanguageModel(dataset.vocab_size)\n\n@pytest.fixture\ndef device():\n return 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"3b2c33bcd73f31217a9236b498c180e79716f55a1fead5651ce753267ac626cf","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.conftest.model","uri":"program://bigram_language_model/function/tests.conftest.model#L15-L16","kind":"function","name":"model","path":"tests/conftest.py","language":"python","start_line":15,"end_line":16,"context_start_line":1,"context_end_line":20,"code":"import pytest\nimport torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\n@pytest.fixture\ndef sample_text():\n return \"Hello, World!\"\n\n@pytest.fixture\ndef dataset(sample_text):\n return TextDataset(sample_text, sequence_length=3)\n\n@pytest.fixture\ndef model(dataset):\n return BigramLanguageModel(dataset.vocab_size)\n\n@pytest.fixture\ndef device():\n return 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"3b2c33bcd73f31217a9236b498c180e79716f55a1fead5651ce753267ac626cf","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.conftest.device","uri":"program://bigram_language_model/function/tests.conftest.device#L19-L20","kind":"function","name":"device","path":"tests/conftest.py","language":"python","start_line":19,"end_line":20,"context_start_line":1,"context_end_line":20,"code":"import pytest\nimport torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\n@pytest.fixture\ndef sample_text():\n return \"Hello, World!\"\n\n@pytest.fixture\ndef dataset(sample_text):\n return TextDataset(sample_text, sequence_length=3)\n\n@pytest.fixture\ndef model(dataset):\n return BigramLanguageModel(dataset.vocab_size)\n\n@pytest.fixture\ndef device():\n return 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"3b2c33bcd73f31217a9236b498c180e79716f55a1fead5651ce753267ac626cf","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_model","uri":"program://bigram_language_model/module/tests.test_model#L1-L42","kind":"module","name":"tests.test_model","path":"tests/test_model.py","language":"python","start_line":1,"end_line":42,"context_start_line":1,"context_end_line":42,"code":"import torch\nimport pytest\nfrom src.model import BigramLanguageModel\n\ndef test_model_initialization(model, dataset):\n assert isinstance(model, BigramLanguageModel)\n assert model.bigram_table.shape == (dataset.vocab_size, dataset.vocab_size)\n\ndef test_model_forward(model, device):\n batch_size, seq_length = 2, 4\n idx = torch.randint(0, model.vocab_size, (batch_size, seq_length)).to(device)\n output = model(idx)\n \n assert output.shape == (batch_size, seq_length, model.vocab_size)\n assert not torch.isnan(output).any()\n\ndef test_model_generate(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n max_new_tokens = 10\n \n generated = model.generate(context, max_new_tokens)\n assert generated.shape == (1, max_new_tokens + 1)\n assert not torch.isnan(generated).any()\n\ndef test_temperature_effect(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Higher temperature should lead to more randomness\n high_temp = model.generate(context, max_new_tokens=100, temperature=2.0)\n low_temp = model.generate(context, max_new_tokens=100, temperature=0.1)\n \n # Convert to sets to compare unique tokens\n high_temp_unique = len(set(high_temp[0].tolist()))\n low_temp_unique = len(set(low_temp[0].tolist()))\n \n assert high_temp_unique >= low_temp_unique \n\ndef test_bigram_properties(model):\n # Test that each row in bigram_table represents a probability distribution after softmax\n probs = torch.softmax(model.bigram_table, dim=1)\n row_sums = probs.sum(dim=1)\n assert torch.allclose(row_sums, torch.ones_like(row_sums)) ","source_hash":"9cae91005efa680298768f1c8b5421c349678740498d61a2ebca50dfdb669c28","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_model.test_model_initialization","uri":"program://bigram_language_model/function/tests.test_model.test_model_initialization#L5-L7","kind":"function","name":"test_model_initialization","path":"tests/test_model.py","language":"python","start_line":5,"end_line":7,"context_start_line":1,"context_end_line":27,"code":"import torch\nimport pytest\nfrom src.model import BigramLanguageModel\n\ndef test_model_initialization(model, dataset):\n assert isinstance(model, BigramLanguageModel)\n assert model.bigram_table.shape == (dataset.vocab_size, dataset.vocab_size)\n\ndef test_model_forward(model, device):\n batch_size, seq_length = 2, 4\n idx = torch.randint(0, model.vocab_size, (batch_size, seq_length)).to(device)\n output = model(idx)\n \n assert output.shape == (batch_size, seq_length, model.vocab_size)\n assert not torch.isnan(output).any()\n\ndef test_model_generate(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n max_new_tokens = 10\n \n generated = model.generate(context, max_new_tokens)\n assert generated.shape == (1, max_new_tokens + 1)\n assert not torch.isnan(generated).any()\n\ndef test_temperature_effect(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n ","source_hash":"9cae91005efa680298768f1c8b5421c349678740498d61a2ebca50dfdb669c28","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_model.test_model_forward","uri":"program://bigram_language_model/function/tests.test_model.test_model_forward#L9-L15","kind":"function","name":"test_model_forward","path":"tests/test_model.py","language":"python","start_line":9,"end_line":15,"context_start_line":1,"context_end_line":35,"code":"import torch\nimport pytest\nfrom src.model import BigramLanguageModel\n\ndef test_model_initialization(model, dataset):\n assert isinstance(model, BigramLanguageModel)\n assert model.bigram_table.shape == (dataset.vocab_size, dataset.vocab_size)\n\ndef test_model_forward(model, device):\n batch_size, seq_length = 2, 4\n idx = torch.randint(0, model.vocab_size, (batch_size, seq_length)).to(device)\n output = model(idx)\n \n assert output.shape == (batch_size, seq_length, model.vocab_size)\n assert not torch.isnan(output).any()\n\ndef test_model_generate(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n max_new_tokens = 10\n \n generated = model.generate(context, max_new_tokens)\n assert generated.shape == (1, max_new_tokens + 1)\n assert not torch.isnan(generated).any()\n\ndef test_temperature_effect(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Higher temperature should lead to more randomness\n high_temp = model.generate(context, max_new_tokens=100, temperature=2.0)\n low_temp = model.generate(context, max_new_tokens=100, temperature=0.1)\n \n # Convert to sets to compare unique tokens\n high_temp_unique = len(set(high_temp[0].tolist()))\n low_temp_unique = len(set(low_temp[0].tolist()))\n ","source_hash":"9cae91005efa680298768f1c8b5421c349678740498d61a2ebca50dfdb669c28","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_model.test_model_generate","uri":"program://bigram_language_model/function/tests.test_model.test_model_generate#L17-L23","kind":"function","name":"test_model_generate","path":"tests/test_model.py","language":"python","start_line":17,"end_line":23,"context_start_line":1,"context_end_line":42,"code":"import torch\nimport pytest\nfrom src.model import BigramLanguageModel\n\ndef test_model_initialization(model, dataset):\n assert isinstance(model, BigramLanguageModel)\n assert model.bigram_table.shape == (dataset.vocab_size, dataset.vocab_size)\n\ndef test_model_forward(model, device):\n batch_size, seq_length = 2, 4\n idx = torch.randint(0, model.vocab_size, (batch_size, seq_length)).to(device)\n output = model(idx)\n \n assert output.shape == (batch_size, seq_length, model.vocab_size)\n assert not torch.isnan(output).any()\n\ndef test_model_generate(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n max_new_tokens = 10\n \n generated = model.generate(context, max_new_tokens)\n assert generated.shape == (1, max_new_tokens + 1)\n assert not torch.isnan(generated).any()\n\ndef test_temperature_effect(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Higher temperature should lead to more randomness\n high_temp = model.generate(context, max_new_tokens=100, temperature=2.0)\n low_temp = model.generate(context, max_new_tokens=100, temperature=0.1)\n \n # Convert to sets to compare unique tokens\n high_temp_unique = len(set(high_temp[0].tolist()))\n low_temp_unique = len(set(low_temp[0].tolist()))\n \n assert high_temp_unique >= low_temp_unique \n\ndef test_bigram_properties(model):\n # Test that each row in bigram_table represents a probability distribution after softmax\n probs = torch.softmax(model.bigram_table, dim=1)\n row_sums = probs.sum(dim=1)\n assert torch.allclose(row_sums, torch.ones_like(row_sums)) ","source_hash":"9cae91005efa680298768f1c8b5421c349678740498d61a2ebca50dfdb669c28","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_model.test_temperature_effect","uri":"program://bigram_language_model/function/tests.test_model.test_temperature_effect#L25-L36","kind":"function","name":"test_temperature_effect","path":"tests/test_model.py","language":"python","start_line":25,"end_line":36,"context_start_line":5,"context_end_line":42,"code":"def test_model_initialization(model, dataset):\n assert isinstance(model, BigramLanguageModel)\n assert model.bigram_table.shape == (dataset.vocab_size, dataset.vocab_size)\n\ndef test_model_forward(model, device):\n batch_size, seq_length = 2, 4\n idx = torch.randint(0, model.vocab_size, (batch_size, seq_length)).to(device)\n output = model(idx)\n \n assert output.shape == (batch_size, seq_length, model.vocab_size)\n assert not torch.isnan(output).any()\n\ndef test_model_generate(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n max_new_tokens = 10\n \n generated = model.generate(context, max_new_tokens)\n assert generated.shape == (1, max_new_tokens + 1)\n assert not torch.isnan(generated).any()\n\ndef test_temperature_effect(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Higher temperature should lead to more randomness\n high_temp = model.generate(context, max_new_tokens=100, temperature=2.0)\n low_temp = model.generate(context, max_new_tokens=100, temperature=0.1)\n \n # Convert to sets to compare unique tokens\n high_temp_unique = len(set(high_temp[0].tolist()))\n low_temp_unique = len(set(low_temp[0].tolist()))\n \n assert high_temp_unique >= low_temp_unique \n\ndef test_bigram_properties(model):\n # Test that each row in bigram_table represents a probability distribution after softmax\n probs = torch.softmax(model.bigram_table, dim=1)\n row_sums = probs.sum(dim=1)\n assert torch.allclose(row_sums, torch.ones_like(row_sums)) ","source_hash":"9cae91005efa680298768f1c8b5421c349678740498d61a2ebca50dfdb669c28","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_model.test_bigram_properties","uri":"program://bigram_language_model/function/tests.test_model.test_bigram_properties#L38-L42","kind":"function","name":"test_bigram_properties","path":"tests/test_model.py","language":"python","start_line":38,"end_line":42,"context_start_line":18,"context_end_line":42,"code":" context = torch.zeros((1, 1), dtype=torch.long, device=device)\n max_new_tokens = 10\n \n generated = model.generate(context, max_new_tokens)\n assert generated.shape == (1, max_new_tokens + 1)\n assert not torch.isnan(generated).any()\n\ndef test_temperature_effect(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Higher temperature should lead to more randomness\n high_temp = model.generate(context, max_new_tokens=100, temperature=2.0)\n low_temp = model.generate(context, max_new_tokens=100, temperature=0.1)\n \n # Convert to sets to compare unique tokens\n high_temp_unique = len(set(high_temp[0].tolist()))\n low_temp_unique = len(set(low_temp[0].tolist()))\n \n assert high_temp_unique >= low_temp_unique \n\ndef test_bigram_properties(model):\n # Test that each row in bigram_table represents a probability distribution after softmax\n probs = torch.softmax(model.bigram_table, dim=1)\n row_sums = probs.sum(dim=1)\n assert torch.allclose(row_sums, torch.ones_like(row_sums)) ","source_hash":"9cae91005efa680298768f1c8b5421c349678740498d61a2ebca50dfdb669c28","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_generation","uri":"program://bigram_language_model/module/tests.test_generation#L1-L33","kind":"module","name":"tests.test_generation","path":"tests/test_generation.py","language":"python","start_line":1,"end_line":33,"context_start_line":1,"context_end_line":33,"code":"import torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\ndef test_end_to_end_generation(sample_text, device):\n # Create dataset and model\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n \n # Generate text\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n generated_indices = model.generate(context, max_new_tokens=20)\n generated_text = dataset.decode(generated_indices[0].tolist())\n \n assert len(generated_text) == 21 # context + max_new_tokens\n assert all(c in dataset.stoi for c in generated_text)\n\ndef test_generation_with_different_temperatures(sample_text, device):\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Generate with different temperatures\n temps = [0.1, 0.5, 1.0, 2.0]\n generated_texts = []\n \n for temp in temps:\n indices = model.generate(context, max_new_tokens=50, temperature=temp)\n generated_texts.append(dataset.decode(indices[0].tolist()))\n \n # Check that all generated texts are valid\n for text in generated_texts:\n assert all(c in dataset.stoi for c in text) ","source_hash":"eff9d30b0618343d5b7ddbf6fe50ae294475a720fe08a0785174d4911372e91c","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_generation.test_end_to_end_generation","uri":"program://bigram_language_model/function/tests.test_generation.test_end_to_end_generation#L5-L16","kind":"function","name":"test_end_to_end_generation","path":"tests/test_generation.py","language":"python","start_line":5,"end_line":16,"context_start_line":1,"context_end_line":33,"code":"import torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\ndef test_end_to_end_generation(sample_text, device):\n # Create dataset and model\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n \n # Generate text\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n generated_indices = model.generate(context, max_new_tokens=20)\n generated_text = dataset.decode(generated_indices[0].tolist())\n \n assert len(generated_text) == 21 # context + max_new_tokens\n assert all(c in dataset.stoi for c in generated_text)\n\ndef test_generation_with_different_temperatures(sample_text, device):\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Generate with different temperatures\n temps = [0.1, 0.5, 1.0, 2.0]\n generated_texts = []\n \n for temp in temps:\n indices = model.generate(context, max_new_tokens=50, temperature=temp)\n generated_texts.append(dataset.decode(indices[0].tolist()))\n \n # Check that all generated texts are valid\n for text in generated_texts:\n assert all(c in dataset.stoi for c in text) ","source_hash":"eff9d30b0618343d5b7ddbf6fe50ae294475a720fe08a0785174d4911372e91c","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_generation.test_generation_with_different_temperatures","uri":"program://bigram_language_model/function/tests.test_generation.test_generation_with_different_temperatures#L18-L33","kind":"function","name":"test_generation_with_different_temperatures","path":"tests/test_generation.py","language":"python","start_line":18,"end_line":33,"context_start_line":1,"context_end_line":33,"code":"import torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\ndef test_end_to_end_generation(sample_text, device):\n # Create dataset and model\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n \n # Generate text\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n generated_indices = model.generate(context, max_new_tokens=20)\n generated_text = dataset.decode(generated_indices[0].tolist())\n \n assert len(generated_text) == 21 # context + max_new_tokens\n assert all(c in dataset.stoi for c in generated_text)\n\ndef test_generation_with_different_temperatures(sample_text, device):\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n \n # Generate with different temperatures\n temps = [0.1, 0.5, 1.0, 2.0]\n generated_texts = []\n \n for temp in temps:\n indices = model.generate(context, max_new_tokens=50, temperature=temp)\n generated_texts.append(dataset.decode(indices[0].tolist()))\n \n # Check that all generated texts are valid\n for text in generated_texts:\n assert all(c in dataset.stoi for c in text) ","source_hash":"eff9d30b0618343d5b7ddbf6fe50ae294475a720fe08a0785174d4911372e91c","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_dataset","uri":"program://bigram_language_model/module/tests.test_dataset#L1-L31","kind":"module","name":"tests.test_dataset","path":"tests/test_dataset.py","language":"python","start_line":1,"end_line":31,"context_start_line":1,"context_end_line":31,"code":"import torch\nimport pytest\nfrom src.data import TextDataset\n\ndef test_dataset_initialization(sample_text):\n dataset = TextDataset(sample_text, sequence_length=3)\n assert dataset.sequence_length == 3\n assert dataset.vocab_size == len(set(sample_text))\n assert len(dataset.stoi) == len(dataset.itos)\n\ndef test_dataset_encoding_decoding(dataset, sample_text):\n # Test if we can encode and decode back to the original text\n indices = [dataset.stoi[ch] for ch in sample_text[:dataset.sequence_length]]\n decoded = dataset.decode(indices)\n assert decoded == sample_text[:dataset.sequence_length]\n\ndef test_dataset_getitem(dataset):\n x, y = dataset[0]\n assert isinstance(x, torch.Tensor)\n assert isinstance(y, torch.Tensor)\n assert x.shape == (dataset.sequence_length,)\n assert y.shape == (dataset.sequence_length,)\n assert (y[:-1] == x[1:]).all() # Check if y is shifted by one position\n\ndef test_dataset_length(dataset, sample_text):\n expected_length = len(sample_text) - dataset.sequence_length\n assert len(dataset) == expected_length\n\ndef test_invalid_sequence_length():\n with pytest.raises(ValueError):\n TextDataset(\"short\", sequence_length=10) ","source_hash":"5f5e08cd5e5848a88da3dcf3205a2c97dfd4e145c6a7b68b8d5b1a2ce730b1fe","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_dataset.test_dataset_initialization","uri":"program://bigram_language_model/function/tests.test_dataset.test_dataset_initialization#L5-L9","kind":"function","name":"test_dataset_initialization","path":"tests/test_dataset.py","language":"python","start_line":5,"end_line":9,"context_start_line":1,"context_end_line":29,"code":"import torch\nimport pytest\nfrom src.data import TextDataset\n\ndef test_dataset_initialization(sample_text):\n dataset = TextDataset(sample_text, sequence_length=3)\n assert dataset.sequence_length == 3\n assert dataset.vocab_size == len(set(sample_text))\n assert len(dataset.stoi) == len(dataset.itos)\n\ndef test_dataset_encoding_decoding(dataset, sample_text):\n # Test if we can encode and decode back to the original text\n indices = [dataset.stoi[ch] for ch in sample_text[:dataset.sequence_length]]\n decoded = dataset.decode(indices)\n assert decoded == sample_text[:dataset.sequence_length]\n\ndef test_dataset_getitem(dataset):\n x, y = dataset[0]\n assert isinstance(x, torch.Tensor)\n assert isinstance(y, torch.Tensor)\n assert x.shape == (dataset.sequence_length,)\n assert y.shape == (dataset.sequence_length,)\n assert (y[:-1] == x[1:]).all() # Check if y is shifted by one position\n\ndef test_dataset_length(dataset, sample_text):\n expected_length = len(sample_text) - dataset.sequence_length\n assert len(dataset) == expected_length\n\ndef test_invalid_sequence_length():","source_hash":"5f5e08cd5e5848a88da3dcf3205a2c97dfd4e145c6a7b68b8d5b1a2ce730b1fe","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_dataset.test_dataset_encoding_decoding","uri":"program://bigram_language_model/function/tests.test_dataset.test_dataset_encoding_decoding#L11-L15","kind":"function","name":"test_dataset_encoding_decoding","path":"tests/test_dataset.py","language":"python","start_line":11,"end_line":15,"context_start_line":1,"context_end_line":31,"code":"import torch\nimport pytest\nfrom src.data import TextDataset\n\ndef test_dataset_initialization(sample_text):\n dataset = TextDataset(sample_text, sequence_length=3)\n assert dataset.sequence_length == 3\n assert dataset.vocab_size == len(set(sample_text))\n assert len(dataset.stoi) == len(dataset.itos)\n\ndef test_dataset_encoding_decoding(dataset, sample_text):\n # Test if we can encode and decode back to the original text\n indices = [dataset.stoi[ch] for ch in sample_text[:dataset.sequence_length]]\n decoded = dataset.decode(indices)\n assert decoded == sample_text[:dataset.sequence_length]\n\ndef test_dataset_getitem(dataset):\n x, y = dataset[0]\n assert isinstance(x, torch.Tensor)\n assert isinstance(y, torch.Tensor)\n assert x.shape == (dataset.sequence_length,)\n assert y.shape == (dataset.sequence_length,)\n assert (y[:-1] == x[1:]).all() # Check if y is shifted by one position\n\ndef test_dataset_length(dataset, sample_text):\n expected_length = len(sample_text) - dataset.sequence_length\n assert len(dataset) == expected_length\n\ndef test_invalid_sequence_length():\n with pytest.raises(ValueError):\n TextDataset(\"short\", sequence_length=10) ","source_hash":"5f5e08cd5e5848a88da3dcf3205a2c97dfd4e145c6a7b68b8d5b1a2ce730b1fe","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_dataset.test_dataset_getitem","uri":"program://bigram_language_model/function/tests.test_dataset.test_dataset_getitem#L17-L23","kind":"function","name":"test_dataset_getitem","path":"tests/test_dataset.py","language":"python","start_line":17,"end_line":23,"context_start_line":1,"context_end_line":31,"code":"import torch\nimport pytest\nfrom src.data import TextDataset\n\ndef test_dataset_initialization(sample_text):\n dataset = TextDataset(sample_text, sequence_length=3)\n assert dataset.sequence_length == 3\n assert dataset.vocab_size == len(set(sample_text))\n assert len(dataset.stoi) == len(dataset.itos)\n\ndef test_dataset_encoding_decoding(dataset, sample_text):\n # Test if we can encode and decode back to the original text\n indices = [dataset.stoi[ch] for ch in sample_text[:dataset.sequence_length]]\n decoded = dataset.decode(indices)\n assert decoded == sample_text[:dataset.sequence_length]\n\ndef test_dataset_getitem(dataset):\n x, y = dataset[0]\n assert isinstance(x, torch.Tensor)\n assert isinstance(y, torch.Tensor)\n assert x.shape == (dataset.sequence_length,)\n assert y.shape == (dataset.sequence_length,)\n assert (y[:-1] == x[1:]).all() # Check if y is shifted by one position\n\ndef test_dataset_length(dataset, sample_text):\n expected_length = len(sample_text) - dataset.sequence_length\n assert len(dataset) == expected_length\n\ndef test_invalid_sequence_length():\n with pytest.raises(ValueError):\n TextDataset(\"short\", sequence_length=10) ","source_hash":"5f5e08cd5e5848a88da3dcf3205a2c97dfd4e145c6a7b68b8d5b1a2ce730b1fe","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_dataset.test_dataset_length","uri":"program://bigram_language_model/function/tests.test_dataset.test_dataset_length#L25-L27","kind":"function","name":"test_dataset_length","path":"tests/test_dataset.py","language":"python","start_line":25,"end_line":27,"context_start_line":5,"context_end_line":31,"code":"def test_dataset_initialization(sample_text):\n dataset = TextDataset(sample_text, sequence_length=3)\n assert dataset.sequence_length == 3\n assert dataset.vocab_size == len(set(sample_text))\n assert len(dataset.stoi) == len(dataset.itos)\n\ndef test_dataset_encoding_decoding(dataset, sample_text):\n # Test if we can encode and decode back to the original text\n indices = [dataset.stoi[ch] for ch in sample_text[:dataset.sequence_length]]\n decoded = dataset.decode(indices)\n assert decoded == sample_text[:dataset.sequence_length]\n\ndef test_dataset_getitem(dataset):\n x, y = dataset[0]\n assert isinstance(x, torch.Tensor)\n assert isinstance(y, torch.Tensor)\n assert x.shape == (dataset.sequence_length,)\n assert y.shape == (dataset.sequence_length,)\n assert (y[:-1] == x[1:]).all() # Check if y is shifted by one position\n\ndef test_dataset_length(dataset, sample_text):\n expected_length = len(sample_text) - dataset.sequence_length\n assert len(dataset) == expected_length\n\ndef test_invalid_sequence_length():\n with pytest.raises(ValueError):\n TextDataset(\"short\", sequence_length=10) ","source_hash":"5f5e08cd5e5848a88da3dcf3205a2c97dfd4e145c6a7b68b8d5b1a2ce730b1fe","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:tests.test_dataset.test_invalid_sequence_length","uri":"program://bigram_language_model/function/tests.test_dataset.test_invalid_sequence_length#L29-L31","kind":"function","name":"test_invalid_sequence_length","path":"tests/test_dataset.py","language":"python","start_line":29,"end_line":31,"context_start_line":9,"context_end_line":31,"code":" assert len(dataset.stoi) == len(dataset.itos)\n\ndef test_dataset_encoding_decoding(dataset, sample_text):\n # Test if we can encode and decode back to the original text\n indices = [dataset.stoi[ch] for ch in sample_text[:dataset.sequence_length]]\n decoded = dataset.decode(indices)\n assert decoded == sample_text[:dataset.sequence_length]\n\ndef test_dataset_getitem(dataset):\n x, y = dataset[0]\n assert isinstance(x, torch.Tensor)\n assert isinstance(y, torch.Tensor)\n assert x.shape == (dataset.sequence_length,)\n assert y.shape == (dataset.sequence_length,)\n assert (y[:-1] == x[1:]).all() # Check if y is shifted by one position\n\ndef test_dataset_length(dataset, sample_text):\n expected_length = len(sample_text) - dataset.sequence_length\n assert len(dataset) == expected_length\n\ndef test_invalid_sequence_length():\n with pytest.raises(ValueError):\n TextDataset(\"short\", sequence_length=10) ","source_hash":"5f5e08cd5e5848a88da3dcf3205a2c97dfd4e145c6a7b68b8d5b1a2ce730b1fe","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.utils.config","uri":"program://bigram_language_model/module/src.utils.config#L1-L21","kind":"module","name":"src.utils.config","path":"src/utils/config.py","language":"python","start_line":1,"end_line":21,"context_start_line":1,"context_end_line":21,"code":"from dataclasses import dataclass\nimport torch\n\n@dataclass\nclass TrainingConfig:\n input_file: str\n output_model: str = 'model.pt'\n sequence_length: int = 8\n batch_size: int = 32\n epochs: int = 50\n learning_rate: float = 0.001\n sample_every: int = 10\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu'\n\n@dataclass\nclass GenerationConfig:\n model_file: str\n input_file: str\n num_tokens: int = 100\n temperature: float = 1.0\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"50447ed6c789796dc1279b1c3ba5fc261495b387c9acd15a3dbfdf823599163d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.utils.config.TrainingConfig","uri":"program://bigram_language_model/class/src.utils.config.TrainingConfig#L5-L13","kind":"class","name":"TrainingConfig","path":"src/utils/config.py","language":"python","start_line":5,"end_line":13,"context_start_line":1,"context_end_line":21,"code":"from dataclasses import dataclass\nimport torch\n\n@dataclass\nclass TrainingConfig:\n input_file: str\n output_model: str = 'model.pt'\n sequence_length: int = 8\n batch_size: int = 32\n epochs: int = 50\n learning_rate: float = 0.001\n sample_every: int = 10\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu'\n\n@dataclass\nclass GenerationConfig:\n model_file: str\n input_file: str\n num_tokens: int = 100\n temperature: float = 1.0\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"50447ed6c789796dc1279b1c3ba5fc261495b387c9acd15a3dbfdf823599163d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.utils.config.GenerationConfig","uri":"program://bigram_language_model/class/src.utils.config.GenerationConfig#L16-L21","kind":"class","name":"GenerationConfig","path":"src/utils/config.py","language":"python","start_line":16,"end_line":21,"context_start_line":1,"context_end_line":21,"code":"from dataclasses import dataclass\nimport torch\n\n@dataclass\nclass TrainingConfig:\n input_file: str\n output_model: str = 'model.pt'\n sequence_length: int = 8\n batch_size: int = 32\n epochs: int = 50\n learning_rate: float = 0.001\n sample_every: int = 10\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu'\n\n@dataclass\nclass GenerationConfig:\n model_file: str\n input_file: str\n num_tokens: int = 100\n temperature: float = 1.0\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"50447ed6c789796dc1279b1c3ba5fc261495b387c9acd15a3dbfdf823599163d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.data.dataset","uri":"program://bigram_language_model/module/src.data.dataset#L1-L43","kind":"module","name":"src.data.dataset","path":"src/data/dataset.py","language":"python","start_line":1,"end_line":43,"context_start_line":1,"context_end_line":43,"code":"import torch\nfrom torch.utils.data import Dataset\nfrom collections import Counter\nimport numpy as np\n\nclass TextDataset(Dataset):\n def __init__(self, text, sequence_length=8):\n if not isinstance(sequence_length, int) or sequence_length <= 0:\n raise ValueError(\"sequence_length must be a positive integer\")\n self.sequence_length = sequence_length\n \n # Create vocabulary\n if not isinstance(text, str) or len(text) == 0:\n raise ValueError(\"text must be a non-empty string\")\n chars = sorted(list(set(text)))\n self.vocab_size = len(chars)\n self.stoi = {ch: i for i, ch in enumerate(chars)}\n self.itos = {i: ch for i, ch in enumerate(chars)}\n \n # Encode text\n data = torch.tensor([self.stoi[ch] for ch in text], dtype=torch.long)\n\n # Validate sequence_length relative to data length\n if len(data) <= sequence_length:\n raise ValueError(\"sequence_length must be less than the length of the text\")\n \n # Create sequences\n self.sequences = []\n for i in range(0, len(data) - sequence_length):\n sequence = data[i:i + sequence_length + 1]\n self.sequences.append(sequence)\n \n def __len__(self):\n return len(self.sequences)\n \n def __getitem__(self, idx):\n sequence = self.sequences[idx]\n x = sequence[:-1]\n y = sequence[1:]\n return x, y\n \n def decode(self, indices):\n return ''.join([self.itos[int(i)] for i in indices]) ","source_hash":"5468e3e0a709fdc8a09a9e97914ec6a2d3b32091403e4f31426460f76c9ffade","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.data.dataset.TextDataset","uri":"program://bigram_language_model/class/src.data.dataset.TextDataset#L6-L43","kind":"class","name":"TextDataset","path":"src/data/dataset.py","language":"python","start_line":6,"end_line":43,"context_start_line":1,"context_end_line":43,"code":"import torch\nfrom torch.utils.data import Dataset\nfrom collections import Counter\nimport numpy as np\n\nclass TextDataset(Dataset):\n def __init__(self, text, sequence_length=8):\n if not isinstance(sequence_length, int) or sequence_length <= 0:\n raise ValueError(\"sequence_length must be a positive integer\")\n self.sequence_length = sequence_length\n \n # Create vocabulary\n if not isinstance(text, str) or len(text) == 0:\n raise ValueError(\"text must be a non-empty string\")\n chars = sorted(list(set(text)))\n self.vocab_size = len(chars)\n self.stoi = {ch: i for i, ch in enumerate(chars)}\n self.itos = {i: ch for i, ch in enumerate(chars)}\n \n # Encode text\n data = torch.tensor([self.stoi[ch] for ch in text], dtype=torch.long)\n\n # Validate sequence_length relative to data length\n if len(data) <= sequence_length:\n raise ValueError(\"sequence_length must be less than the length of the text\")\n \n # Create sequences\n self.sequences = []\n for i in range(0, len(data) - sequence_length):\n sequence = data[i:i + sequence_length + 1]\n self.sequences.append(sequence)\n \n def __len__(self):\n return len(self.sequences)\n \n def __getitem__(self, idx):\n sequence = self.sequences[idx]\n x = sequence[:-1]\n y = sequence[1:]\n return x, y\n \n def decode(self, indices):\n return ''.join([self.itos[int(i)] for i in indices]) ","source_hash":"5468e3e0a709fdc8a09a9e97914ec6a2d3b32091403e4f31426460f76c9ffade","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.data.dataset.__init__","uri":"program://bigram_language_model/function/src.data.dataset.__init__#L7-L31","kind":"function","name":"__init__","path":"src/data/dataset.py","language":"python","start_line":7,"end_line":31,"context_start_line":1,"context_end_line":43,"code":"import torch\nfrom torch.utils.data import Dataset\nfrom collections import Counter\nimport numpy as np\n\nclass TextDataset(Dataset):\n def __init__(self, text, sequence_length=8):\n if not isinstance(sequence_length, int) or sequence_length <= 0:\n raise ValueError(\"sequence_length must be a positive integer\")\n self.sequence_length = sequence_length\n \n # Create vocabulary\n if not isinstance(text, str) or len(text) == 0:\n raise ValueError(\"text must be a non-empty string\")\n chars = sorted(list(set(text)))\n self.vocab_size = len(chars)\n self.stoi = {ch: i for i, ch in enumerate(chars)}\n self.itos = {i: ch for i, ch in enumerate(chars)}\n \n # Encode text\n data = torch.tensor([self.stoi[ch] for ch in text], dtype=torch.long)\n\n # Validate sequence_length relative to data length\n if len(data) <= sequence_length:\n raise ValueError(\"sequence_length must be less than the length of the text\")\n \n # Create sequences\n self.sequences = []\n for i in range(0, len(data) - sequence_length):\n sequence = data[i:i + sequence_length + 1]\n self.sequences.append(sequence)\n \n def __len__(self):\n return len(self.sequences)\n \n def __getitem__(self, idx):\n sequence = self.sequences[idx]\n x = sequence[:-1]\n y = sequence[1:]\n return x, y\n \n def decode(self, indices):\n return ''.join([self.itos[int(i)] for i in indices]) ","source_hash":"5468e3e0a709fdc8a09a9e97914ec6a2d3b32091403e4f31426460f76c9ffade","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.data.dataset.__len__","uri":"program://bigram_language_model/function/src.data.dataset.__len__#L33-L34","kind":"function","name":"__len__","path":"src/data/dataset.py","language":"python","start_line":33,"end_line":34,"context_start_line":13,"context_end_line":43,"code":" if not isinstance(text, str) or len(text) == 0:\n raise ValueError(\"text must be a non-empty string\")\n chars = sorted(list(set(text)))\n self.vocab_size = len(chars)\n self.stoi = {ch: i for i, ch in enumerate(chars)}\n self.itos = {i: ch for i, ch in enumerate(chars)}\n \n # Encode text\n data = torch.tensor([self.stoi[ch] for ch in text], dtype=torch.long)\n\n # Validate sequence_length relative to data length\n if len(data) <= sequence_length:\n raise ValueError(\"sequence_length must be less than the length of the text\")\n \n # Create sequences\n self.sequences = []\n for i in range(0, len(data) - sequence_length):\n sequence = data[i:i + sequence_length + 1]\n self.sequences.append(sequence)\n \n def __len__(self):\n return len(self.sequences)\n \n def __getitem__(self, idx):\n sequence = self.sequences[idx]\n x = sequence[:-1]\n y = sequence[1:]\n return x, y\n \n def decode(self, indices):\n return ''.join([self.itos[int(i)] for i in indices]) ","source_hash":"5468e3e0a709fdc8a09a9e97914ec6a2d3b32091403e4f31426460f76c9ffade","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.data.dataset.__getitem__","uri":"program://bigram_language_model/function/src.data.dataset.__getitem__#L36-L40","kind":"function","name":"__getitem__","path":"src/data/dataset.py","language":"python","start_line":36,"end_line":40,"context_start_line":16,"context_end_line":43,"code":" self.vocab_size = len(chars)\n self.stoi = {ch: i for i, ch in enumerate(chars)}\n self.itos = {i: ch for i, ch in enumerate(chars)}\n \n # Encode text\n data = torch.tensor([self.stoi[ch] for ch in text], dtype=torch.long)\n\n # Validate sequence_length relative to data length\n if len(data) <= sequence_length:\n raise ValueError(\"sequence_length must be less than the length of the text\")\n \n # Create sequences\n self.sequences = []\n for i in range(0, len(data) - sequence_length):\n sequence = data[i:i + sequence_length + 1]\n self.sequences.append(sequence)\n \n def __len__(self):\n return len(self.sequences)\n \n def __getitem__(self, idx):\n sequence = self.sequences[idx]\n x = sequence[:-1]\n y = sequence[1:]\n return x, y\n \n def decode(self, indices):\n return ''.join([self.itos[int(i)] for i in indices]) ","source_hash":"5468e3e0a709fdc8a09a9e97914ec6a2d3b32091403e4f31426460f76c9ffade","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.data.dataset.decode","uri":"program://bigram_language_model/function/src.data.dataset.decode#L42-L43","kind":"function","name":"decode","path":"src/data/dataset.py","language":"python","start_line":42,"end_line":43,"context_start_line":22,"context_end_line":43,"code":"\n # Validate sequence_length relative to data length\n if len(data) <= sequence_length:\n raise ValueError(\"sequence_length must be less than the length of the text\")\n \n # Create sequences\n self.sequences = []\n for i in range(0, len(data) - sequence_length):\n sequence = data[i:i + sequence_length + 1]\n self.sequences.append(sequence)\n \n def __len__(self):\n return len(self.sequences)\n \n def __getitem__(self, idx):\n sequence = self.sequences[idx]\n x = sequence[:-1]\n y = sequence[1:]\n return x, y\n \n def decode(self, indices):\n return ''.join([self.itos[int(i)] for i in indices]) ","source_hash":"5468e3e0a709fdc8a09a9e97914ec6a2d3b32091403e4f31426460f76c9ffade","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.model.bigram_model","uri":"program://bigram_language_model/module/src.model.bigram_model#L1-L100","kind":"module","name":"src.model.bigram_model","path":"src/model/bigram_model.py","language":"python","start_line":1,"end_line":100,"context_start_line":1,"context_end_line":100,"code":"import torch\nimport torch.nn as nn\n\nclass BigramLanguageModel(nn.Module):\n \"\"\"A bigram language model implementation.\n \n A bigram language model is one of the simplest forms of language models that predicts\n the probability of a token based solely on the previous token. For a sequence of tokens\n [w₁, w₂, ..., wₙ], the bigram model approximates the probability as:\n \n P(w₁, w₂, ..., wₙ) ≈ P(w₁) * P(w₂|w₁) * P(w₃|w₂) * ... * P(wₙ|wₙ₋₁)\n \n The model maintains a transition matrix (bigram_table) where:\n - Each row i represents the current token\n - Each column j represents the next token\n - Entry (i,j) contains the logit for P(token_j|token_i)\n \n Args:\n vocab_size (int): Size of the vocabulary\n \n Attributes:\n bigram_table (nn.Parameter): Transition matrix of shape (vocab_size, vocab_size)\n containing the log-probabilities of each token following another token\n vocab_size (int): Size of the vocabulary\n \"\"\"\n \n def __init__(self, vocab_size):\n super().__init__()\n # Initialize transition matrix with zeros\n # After training, bigram_table[i,j] will contain the log-probability\n # of token j following token i\n self.bigram_table = nn.Parameter(torch.zeros(vocab_size, vocab_size))\n self.vocab_size = vocab_size\n \n def forward(self, idx):\n \"\"\"Compute the bigram probabilities for a batch of sequences.\n \n For each position in the input sequence, computes logits for the probability\n distribution over the next token.\n \n Args:\n idx (torch.Tensor): Batch of token sequences of shape (batch_size, sequence_length)\n \n Returns:\n torch.Tensor: Logits for next-token predictions of shape \n (batch_size, sequence_length, vocab_size)\n \"\"\"\n B, T = idx.shape\n \n # Index into bigram_table to get next-token logits for each position\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n # Select rows for each token id efficiently and reshape\n rows = table.index_select(0, idx.view(-1)) # (B*T, vocab_size)\n logits = rows.view(B, T, self.vocab_size)\n \n return logits\n \n def generate(self, idx, max_new_tokens, temperature=1.0):\n \"\"\"Generate new tokens autoregressively using the bigram probabilities.\n \n Starting from the last token in the provided context, generates new tokens\n one at a time by:\n 1. Getting the probability distribution for the next token\n 2. Applying temperature scaling to control randomness\n 3. Sampling from the resulting distribution\n \n Args:\n idx (torch.Tensor): Starting context of shape (batch_size, sequence_length)\n max_new_tokens (int): Number of new tokens to generate\n temperature (float, optional): Temperature for sampling. Higher values make the\n distribution more uniform, lower values make it more peaky. Defaults to 1.0.\n \n Returns:\n torch.Tensor: Generated sequence including the context,\n shape (batch_size, sequence_length + max_new_tokens)\n \"\"\"\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n\n for _ in range(max_new_tokens):\n # Only need the last token for bigram prediction\n last_token = idx[:, -1:] # (B, 1)\n rows = table.index_select(0, last_token.view(-1)) # (B*1, vocab_size)\n logits = rows.view(idx.size(0), 1, self.vocab_size) # (B, 1, vocab_size)\n \n # Apply temperature scaling to logits\n # Higher temperature = more random, Lower = more deterministic\n logits = logits[:, -1, :] / temperature # (B, vocab_size)\n \n # Convert logits to probabilities\n probs = torch.softmax(logits, dim=-1) # (B, vocab_size)\n \n # Sample next token from the probability distribution\n idx_next = torch.multinomial(probs, num_samples=1) # (B, 1)\n \n # Append new token to the sequence\n idx = torch.cat((idx, idx_next), dim=1) # (B, T+1)\n \n return idx ","source_hash":"7f08fa5371181a2c0680582f47c9bc2369d539931a65d6d76e5bb067419b565d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.model.bigram_model.BigramLanguageModel","uri":"program://bigram_language_model/class/src.model.bigram_model.BigramLanguageModel#L4-L100","kind":"class","name":"BigramLanguageModel","path":"src/model/bigram_model.py","language":"python","start_line":4,"end_line":100,"context_start_line":1,"context_end_line":100,"code":"import torch\nimport torch.nn as nn\n\nclass BigramLanguageModel(nn.Module):\n \"\"\"A bigram language model implementation.\n \n A bigram language model is one of the simplest forms of language models that predicts\n the probability of a token based solely on the previous token. For a sequence of tokens\n [w₁, w₂, ..., wₙ], the bigram model approximates the probability as:\n \n P(w₁, w₂, ..., wₙ) ≈ P(w₁) * P(w₂|w₁) * P(w₃|w₂) * ... * P(wₙ|wₙ₋₁)\n \n The model maintains a transition matrix (bigram_table) where:\n - Each row i represents the current token\n - Each column j represents the next token\n - Entry (i,j) contains the logit for P(token_j|token_i)\n \n Args:\n vocab_size (int): Size of the vocabulary\n \n Attributes:\n bigram_table (nn.Parameter): Transition matrix of shape (vocab_size, vocab_size)\n containing the log-probabilities of each token following another token\n vocab_size (int): Size of the vocabulary\n \"\"\"\n \n def __init__(self, vocab_size):\n super().__init__()\n # Initialize transition matrix with zeros\n # After training, bigram_table[i,j] will contain the log-probability\n # of token j following token i\n self.bigram_table = nn.Parameter(torch.zeros(vocab_size, vocab_size))\n self.vocab_size = vocab_size\n \n def forward(self, idx):\n \"\"\"Compute the bigram probabilities for a batch of sequences.\n \n For each position in the input sequence, computes logits for the probability\n distribution over the next token.\n \n Args:\n idx (torch.Tensor): Batch of token sequences of shape (batch_size, sequence_length)\n \n Returns:\n torch.Tensor: Logits for next-token predictions of shape \n (batch_size, sequence_length, vocab_size)\n \"\"\"\n B, T = idx.shape\n \n # Index into bigram_table to get next-token logits for each position\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n # Select rows for each token id efficiently and reshape\n rows = table.index_select(0, idx.view(-1)) # (B*T, vocab_size)\n logits = rows.view(B, T, self.vocab_size)\n \n return logits\n \n def generate(self, idx, max_new_tokens, temperature=1.0):\n \"\"\"Generate new tokens autoregressively using the bigram probabilities.\n \n Starting from the last token in the provided context, generates new tokens\n one at a time by:\n 1. Getting the probability distribution for the next token\n 2. Applying temperature scaling to control randomness\n 3. Sampling from the resulting distribution\n \n Args:\n idx (torch.Tensor): Starting context of shape (batch_size, sequence_length)\n max_new_tokens (int): Number of new tokens to generate\n temperature (float, optional): Temperature for sampling. Higher values make the\n distribution more uniform, lower values make it more peaky. Defaults to 1.0.\n \n Returns:\n torch.Tensor: Generated sequence including the context,\n shape (batch_size, sequence_length + max_new_tokens)\n \"\"\"\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n\n for _ in range(max_new_tokens):\n # Only need the last token for bigram prediction\n last_token = idx[:, -1:] # (B, 1)\n rows = table.index_select(0, last_token.view(-1)) # (B*1, vocab_size)\n logits = rows.view(idx.size(0), 1, self.vocab_size) # (B, 1, vocab_size)\n \n # Apply temperature scaling to logits\n # Higher temperature = more random, Lower = more deterministic\n logits = logits[:, -1, :] / temperature # (B, vocab_size)\n \n # Convert logits to probabilities\n probs = torch.softmax(logits, dim=-1) # (B, vocab_size)\n \n # Sample next token from the probability distribution\n idx_next = torch.multinomial(probs, num_samples=1) # (B, 1)\n \n # Append new token to the sequence\n idx = torch.cat((idx, idx_next), dim=1) # (B, T+1)\n \n return idx ","source_hash":"7f08fa5371181a2c0680582f47c9bc2369d539931a65d6d76e5bb067419b565d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.model.bigram_model.__init__","uri":"program://bigram_language_model/function/src.model.bigram_model.__init__#L27-L33","kind":"function","name":"__init__","path":"src/model/bigram_model.py","language":"python","start_line":27,"end_line":33,"context_start_line":7,"context_end_line":53,"code":" A bigram language model is one of the simplest forms of language models that predicts\n the probability of a token based solely on the previous token. For a sequence of tokens\n [w₁, w₂, ..., wₙ], the bigram model approximates the probability as:\n \n P(w₁, w₂, ..., wₙ) ≈ P(w₁) * P(w₂|w₁) * P(w₃|w₂) * ... * P(wₙ|wₙ₋₁)\n \n The model maintains a transition matrix (bigram_table) where:\n - Each row i represents the current token\n - Each column j represents the next token\n - Entry (i,j) contains the logit for P(token_j|token_i)\n \n Args:\n vocab_size (int): Size of the vocabulary\n \n Attributes:\n bigram_table (nn.Parameter): Transition matrix of shape (vocab_size, vocab_size)\n containing the log-probabilities of each token following another token\n vocab_size (int): Size of the vocabulary\n \"\"\"\n \n def __init__(self, vocab_size):\n super().__init__()\n # Initialize transition matrix with zeros\n # After training, bigram_table[i,j] will contain the log-probability\n # of token j following token i\n self.bigram_table = nn.Parameter(torch.zeros(vocab_size, vocab_size))\n self.vocab_size = vocab_size\n \n def forward(self, idx):\n \"\"\"Compute the bigram probabilities for a batch of sequences.\n \n For each position in the input sequence, computes logits for the probability\n distribution over the next token.\n \n Args:\n idx (torch.Tensor): Batch of token sequences of shape (batch_size, sequence_length)\n \n Returns:\n torch.Tensor: Logits for next-token predictions of shape \n (batch_size, sequence_length, vocab_size)\n \"\"\"\n B, T = idx.shape\n \n # Index into bigram_table to get next-token logits for each position\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n # Select rows for each token id efficiently and reshape","source_hash":"7f08fa5371181a2c0680582f47c9bc2369d539931a65d6d76e5bb067419b565d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.model.bigram_model.forward","uri":"program://bigram_language_model/function/src.model.bigram_model.forward#L35-L57","kind":"function","name":"forward","path":"src/model/bigram_model.py","language":"python","start_line":35,"end_line":57,"context_start_line":15,"context_end_line":77,"code":" - Each column j represents the next token\n - Entry (i,j) contains the logit for P(token_j|token_i)\n \n Args:\n vocab_size (int): Size of the vocabulary\n \n Attributes:\n bigram_table (nn.Parameter): Transition matrix of shape (vocab_size, vocab_size)\n containing the log-probabilities of each token following another token\n vocab_size (int): Size of the vocabulary\n \"\"\"\n \n def __init__(self, vocab_size):\n super().__init__()\n # Initialize transition matrix with zeros\n # After training, bigram_table[i,j] will contain the log-probability\n # of token j following token i\n self.bigram_table = nn.Parameter(torch.zeros(vocab_size, vocab_size))\n self.vocab_size = vocab_size\n \n def forward(self, idx):\n \"\"\"Compute the bigram probabilities for a batch of sequences.\n \n For each position in the input sequence, computes logits for the probability\n distribution over the next token.\n \n Args:\n idx (torch.Tensor): Batch of token sequences of shape (batch_size, sequence_length)\n \n Returns:\n torch.Tensor: Logits for next-token predictions of shape \n (batch_size, sequence_length, vocab_size)\n \"\"\"\n B, T = idx.shape\n \n # Index into bigram_table to get next-token logits for each position\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n # Select rows for each token id efficiently and reshape\n rows = table.index_select(0, idx.view(-1)) # (B*T, vocab_size)\n logits = rows.view(B, T, self.vocab_size)\n \n return logits\n \n def generate(self, idx, max_new_tokens, temperature=1.0):\n \"\"\"Generate new tokens autoregressively using the bigram probabilities.\n \n Starting from the last token in the provided context, generates new tokens\n one at a time by:\n 1. Getting the probability distribution for the next token\n 2. Applying temperature scaling to control randomness\n 3. Sampling from the resulting distribution\n \n Args:\n idx (torch.Tensor): Starting context of shape (batch_size, sequence_length)\n max_new_tokens (int): Number of new tokens to generate\n temperature (float, optional): Temperature for sampling. Higher values make the\n distribution more uniform, lower values make it more peaky. Defaults to 1.0.\n \n Returns:\n torch.Tensor: Generated sequence including the context,\n shape (batch_size, sequence_length + max_new_tokens)\n \"\"\"","source_hash":"7f08fa5371181a2c0680582f47c9bc2369d539931a65d6d76e5bb067419b565d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:src.model.bigram_model.generate","uri":"program://bigram_language_model/function/src.model.bigram_model.generate#L59-L100","kind":"function","name":"generate","path":"src/model/bigram_model.py","language":"python","start_line":59,"end_line":100,"context_start_line":39,"context_end_line":100,"code":" distribution over the next token.\n \n Args:\n idx (torch.Tensor): Batch of token sequences of shape (batch_size, sequence_length)\n \n Returns:\n torch.Tensor: Logits for next-token predictions of shape \n (batch_size, sequence_length, vocab_size)\n \"\"\"\n B, T = idx.shape\n \n # Index into bigram_table to get next-token logits for each position\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n # Select rows for each token id efficiently and reshape\n rows = table.index_select(0, idx.view(-1)) # (B*T, vocab_size)\n logits = rows.view(B, T, self.vocab_size)\n \n return logits\n \n def generate(self, idx, max_new_tokens, temperature=1.0):\n \"\"\"Generate new tokens autoregressively using the bigram probabilities.\n \n Starting from the last token in the provided context, generates new tokens\n one at a time by:\n 1. Getting the probability distribution for the next token\n 2. Applying temperature scaling to control randomness\n 3. Sampling from the resulting distribution\n \n Args:\n idx (torch.Tensor): Starting context of shape (batch_size, sequence_length)\n max_new_tokens (int): Number of new tokens to generate\n temperature (float, optional): Temperature for sampling. Higher values make the\n distribution more uniform, lower values make it more peaky. Defaults to 1.0.\n \n Returns:\n torch.Tensor: Generated sequence including the context,\n shape (batch_size, sequence_length + max_new_tokens)\n \"\"\"\n # Ensure the parameter is on the same device as the indices\n table = self.bigram_table.to(idx.device)\n\n for _ in range(max_new_tokens):\n # Only need the last token for bigram prediction\n last_token = idx[:, -1:] # (B, 1)\n rows = table.index_select(0, last_token.view(-1)) # (B*1, vocab_size)\n logits = rows.view(idx.size(0), 1, self.vocab_size) # (B, 1, vocab_size)\n \n # Apply temperature scaling to logits\n # Higher temperature = more random, Lower = more deterministic\n logits = logits[:, -1, :] / temperature # (B, vocab_size)\n \n # Convert logits to probabilities\n probs = torch.softmax(logits, dim=-1) # (B, vocab_size)\n \n # Sample next token from the probability distribution\n idx_next = torch.multinomial(probs, num_samples=1) # (B, 1)\n \n # Append new token to the sequence\n idx = torch.cat((idx, idx_next), dim=1) # (B, T+1)\n \n return idx ","source_hash":"7f08fa5371181a2c0680582f47c9bc2369d539931a65d6d76e5bb067419b565d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:scripts.generate","uri":"program://bigram_language_model/module/scripts.generate#L1-L3","kind":"module","name":"scripts.generate","path":"scripts/generate.py","language":"python","start_line":1,"end_line":3,"context_start_line":1,"context_end_line":3,"code":"from src.model import BigramLanguageModel\nfrom src.data import TextDataset\nfrom src.utils.config import GenerationConfig ","source_hash":"90ac3dff3ff7aa8afcae01234703f4d5129e5a3d7e1624fbdc676359741f519e","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"py:scripts.train","uri":"program://bigram_language_model/module/scripts.train#L1-L3","kind":"module","name":"scripts.train","path":"scripts/train.py","language":"python","start_line":1,"end_line":3,"context_start_line":1,"context_end_line":3,"code":"from src.model import BigramLanguageModel\nfrom src.data import TextDataset\nfrom src.utils.config import TrainingConfig ","source_hash":"40d529ee6956424de744074ea1718787e0803862dd6765919dbf185ca3efaf05","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:setup.py","uri":"program://bigram_language_model/file/setup.py","kind":"file","name":"setup.py","path":"setup.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"from setuptools import setup, find_packages\n\nsetup(\n name=\"bigram-lm\",\n version=\"0.1.0\",\n packages=find_packages(),\n install_requires=[\n \"torch>=2.0.0\",\n \"numpy>=1.21.0\",\n ],\n extras_require={\n 'test': [\n 'pytest>=7.0.0',\n 'pytest-cov>=4.0.0',\n ],\n },\n author=\"Your Name\",\n author_email=\"your.email@example.com\",\n description=\"A simple bigram language model implementation\",\n long_description=open(\"README.md\").read(),\n long_description_content_type=\"text/markdown\",","source_hash":"bbd5fea0a69a2f9d71ead2e60fbb59c7d795c3a4bc1d6719275a61598e8205ef","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:tests/conftest.py","uri":"program://bigram_language_model/file/tests/conftest.py","kind":"file","name":"tests/conftest.py","path":"tests/conftest.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":20,"code":"import pytest\nimport torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\n@pytest.fixture\ndef sample_text():\n return \"Hello, World!\"\n\n@pytest.fixture\ndef dataset(sample_text):\n return TextDataset(sample_text, sequence_length=3)\n\n@pytest.fixture\ndef model(dataset):\n return BigramLanguageModel(dataset.vocab_size)\n\n@pytest.fixture\ndef device():\n return 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"3b2c33bcd73f31217a9236b498c180e79716f55a1fead5651ce753267ac626cf","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:tests/test_model.py","uri":"program://bigram_language_model/file/tests/test_model.py","kind":"file","name":"tests/test_model.py","path":"tests/test_model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport pytest\nfrom src.model import BigramLanguageModel\n\ndef test_model_initialization(model, dataset):\n assert isinstance(model, BigramLanguageModel)\n assert model.bigram_table.shape == (dataset.vocab_size, dataset.vocab_size)\n\ndef test_model_forward(model, device):\n batch_size, seq_length = 2, 4\n idx = torch.randint(0, model.vocab_size, (batch_size, seq_length)).to(device)\n output = model(idx)\n \n assert output.shape == (batch_size, seq_length, model.vocab_size)\n assert not torch.isnan(output).any()\n\ndef test_model_generate(model, device):\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n max_new_tokens = 10\n \n generated = model.generate(context, max_new_tokens)","source_hash":"9cae91005efa680298768f1c8b5421c349678740498d61a2ebca50dfdb669c28","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:tests/test_generation.py","uri":"program://bigram_language_model/file/tests/test_generation.py","kind":"file","name":"tests/test_generation.py","path":"tests/test_generation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nfrom src.model import BigramLanguageModel\nfrom src.data import TextDataset\n\ndef test_end_to_end_generation(sample_text, device):\n # Create dataset and model\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n \n # Generate text\n context = torch.zeros((1, 1), dtype=torch.long, device=device)\n generated_indices = model.generate(context, max_new_tokens=20)\n generated_text = dataset.decode(generated_indices[0].tolist())\n \n assert len(generated_text) == 21 # context + max_new_tokens\n assert all(c in dataset.stoi for c in generated_text)\n\ndef test_generation_with_different_temperatures(sample_text, device):\n dataset = TextDataset(sample_text, sequence_length=3)\n model = BigramLanguageModel(dataset.vocab_size).to(device)\n context = torch.zeros((1, 1), dtype=torch.long, device=device)","source_hash":"eff9d30b0618343d5b7ddbf6fe50ae294475a720fe08a0785174d4911372e91c","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:tests/test_dataset.py","uri":"program://bigram_language_model/file/tests/test_dataset.py","kind":"file","name":"tests/test_dataset.py","path":"tests/test_dataset.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport pytest\nfrom src.data import TextDataset\n\ndef test_dataset_initialization(sample_text):\n dataset = TextDataset(sample_text, sequence_length=3)\n assert dataset.sequence_length == 3\n assert dataset.vocab_size == len(set(sample_text))\n assert len(dataset.stoi) == len(dataset.itos)\n\ndef test_dataset_encoding_decoding(dataset, sample_text):\n # Test if we can encode and decode back to the original text\n indices = [dataset.stoi[ch] for ch in sample_text[:dataset.sequence_length]]\n decoded = dataset.decode(indices)\n assert decoded == sample_text[:dataset.sequence_length]\n\ndef test_dataset_getitem(dataset):\n x, y = dataset[0]\n assert isinstance(x, torch.Tensor)\n assert isinstance(y, torch.Tensor)\n assert x.shape == (dataset.sequence_length,)","source_hash":"5f5e08cd5e5848a88da3dcf3205a2c97dfd4e145c6a7b68b8d5b1a2ce730b1fe","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:src/__init__.py","uri":"program://bigram_language_model/file/src/__init__.py","kind":"file","name":"src/__init__.py","path":"src/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":1,"code":"# Empty file to make the directory a Python package ","source_hash":"7fd80665840f8fee8a6aadbc88f4fb6bfd5d6b0871edba83b851cafb1d8dd454","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:src/utils/config.py","uri":"program://bigram_language_model/file/src/utils/config.py","kind":"file","name":"src/utils/config.py","path":"src/utils/config.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"from dataclasses import dataclass\nimport torch\n\n@dataclass\nclass TrainingConfig:\n input_file: str\n output_model: str = 'model.pt'\n sequence_length: int = 8\n batch_size: int = 32\n epochs: int = 50\n learning_rate: float = 0.001\n sample_every: int = 10\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu'\n\n@dataclass\nclass GenerationConfig:\n model_file: str\n input_file: str\n num_tokens: int = 100\n temperature: float = 1.0\n device: str = 'cuda' if torch.cuda.is_available() else 'cpu' ","source_hash":"50447ed6c789796dc1279b1c3ba5fc261495b387c9acd15a3dbfdf823599163d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:src/data/dataset.py","uri":"program://bigram_language_model/file/src/data/dataset.py","kind":"file","name":"src/data/dataset.py","path":"src/data/dataset.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nfrom torch.utils.data import Dataset\nfrom collections import Counter\nimport numpy as np\n\nclass TextDataset(Dataset):\n def __init__(self, text, sequence_length=8):\n if not isinstance(sequence_length, int) or sequence_length <= 0:\n raise ValueError(\"sequence_length must be a positive integer\")\n self.sequence_length = sequence_length\n \n # Create vocabulary\n if not isinstance(text, str) or len(text) == 0:\n raise ValueError(\"text must be a non-empty string\")\n chars = sorted(list(set(text)))\n self.vocab_size = len(chars)\n self.stoi = {ch: i for i, ch in enumerate(chars)}\n self.itos = {i: ch for i, ch in enumerate(chars)}\n \n # Encode text\n data = torch.tensor([self.stoi[ch] for ch in text], dtype=torch.long)","source_hash":"5468e3e0a709fdc8a09a9e97914ec6a2d3b32091403e4f31426460f76c9ffade","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:src/data/__init__.py","uri":"program://bigram_language_model/file/src/data/__init__.py","kind":"file","name":"src/data/__init__.py","path":"src/data/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":3,"code":"from .dataset import TextDataset\n\n__all__ = ['TextDataset'] ","source_hash":"2d0721d7ced8b0efee1cb39ec80bd87f9254c10ecc749d2c7d7d48fbac3a02ab","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:src/model/bigram_model.py","uri":"program://bigram_language_model/file/src/model/bigram_model.py","kind":"file","name":"src/model/bigram_model.py","path":"src/model/bigram_model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport torch.nn as nn\n\nclass BigramLanguageModel(nn.Module):\n \"\"\"A bigram language model implementation.\n \n A bigram language model is one of the simplest forms of language models that predicts\n the probability of a token based solely on the previous token. For a sequence of tokens\n [w₁, w₂, ..., wₙ], the bigram model approximates the probability as:\n \n P(w₁, w₂, ..., wₙ) ≈ P(w₁) * P(w₂|w₁) * P(w₃|w₂) * ... * P(wₙ|wₙ₋₁)\n \n The model maintains a transition matrix (bigram_table) where:\n - Each row i represents the current token\n - Each column j represents the next token\n - Entry (i,j) contains the logit for P(token_j|token_i)\n \n Args:\n vocab_size (int): Size of the vocabulary\n \n Attributes:","source_hash":"7f08fa5371181a2c0680582f47c9bc2369d539931a65d6d76e5bb067419b565d","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:src/model/__init__.py","uri":"program://bigram_language_model/file/src/model/__init__.py","kind":"file","name":"src/model/__init__.py","path":"src/model/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":3,"code":"from .bigram_model import BigramLanguageModel\n\n__all__ = ['BigramLanguageModel'] ","source_hash":"6fb7de90ea26e5aab0bf49323d80402af4b10e0b7e19255122718f244bbaa3fd","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:scripts/generate.py","uri":"program://bigram_language_model/file/scripts/generate.py","kind":"file","name":"scripts/generate.py","path":"scripts/generate.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":3,"code":"from src.model import BigramLanguageModel\nfrom src.data import TextDataset\nfrom src.utils.config import GenerationConfig ","source_hash":"90ac3dff3ff7aa8afcae01234703f4d5129e5a3d7e1624fbdc676359741f519e","truncated":false} |
| {"repo_id":"bigram_language_model","entity_id":"file:scripts/train.py","uri":"program://bigram_language_model/file/scripts/train.py","kind":"file","name":"scripts/train.py","path":"scripts/train.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":3,"code":"from src.model import BigramLanguageModel\nfrom src.data import TextDataset\nfrom src.utils.config import TrainingConfig ","source_hash":"40d529ee6956424de744074ea1718787e0803862dd6765919dbf185ca3efaf05","truncated":false} |
|
|
