Python scrapbook.read_notebook方法代码示例

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_text_classification_unified_information(notebooks, tmp):
    notebook_path = notebooks["tc_unified_information"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            DATA_FOLDER=tmp,
            BERT_CACHE_DIR=tmp,
            BATCH_SIZE=32,
            BATCH_SIZE_PRED=512,
            NUM_EPOCHS=1,
            TEST=True,
            QUICK_RUN=True,
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["accuracy"], 0.93, abs=ABS_TOL)
    assert pytest.approx(result["precision"], 0.93, abs=ABS_TOL)
    assert pytest.approx(result["recall"], 0.93, abs=ABS_TOL)
    assert pytest.approx(result["f1"], 0.93, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_text_classification_introspective_rationale(notebooks, tmp):
    notebook_path = notebooks["tc_introspective_rationale"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            DATA_FOLDER=tmp,
            CUDA=torch.cuda.is_available(),
            QUICK_RUN=False,
            MODEL_SAVE_DIR=tmp
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    print(result)
    assert pytest.approx(result["accuracy"], 0.72, abs=ABS_TOL)
    assert pytest.approx(result["anti_accuracy"], 0.69, abs=ABS_TOL)
    assert pytest.approx(result["sparsity"], 0.17, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def assay_one_notebook(notebook_name, test_values):
    """Test a single notebook.

    This uses nbformat to append `nteract-scrapbook` commands to the
    specified notebook. The content of the commands and their expected
    values are stored in the `test_values` dictionary. The keys of this
    dictionary are strings to be used as scrapbook keys. They corresponding
    value is a `ScrapSpec` tuple. The `code` member of this tuple is
    the code (as a string) to be run to generate the scrapbook value. The
    `expected` member is a Python object which is checked for equality with
    the scrapbook value

    Makes certain assumptions about directory layout.
    """
    input_notebook = "notebooks/" + notebook_name + ".ipynb"
    processed_notebook = "./test/notebooks/" + notebook_name + ".processed.ipynb"
    output_notebook = "./test/notebooks/" + notebook_name + ".output.ipynb"

    append_scrapbook_commands(input_notebook, processed_notebook, test_values)
    pm.execute_notebook(processed_notebook, output_notebook)
    nb = sb.read_notebook(output_notebook)

    for k, v in test_values.items():
        assert nb.scraps[k].data == v.expected 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_unilm_abstractive_summarization(notebooks, tmp):
    notebook_path = notebooks["unilm_abstractive_summarization"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            QUICK_RUN=True,
            NUM_GPUS=torch.cuda.device_count(),
            TOP_N=100,
            WARMUP_STEPS=5,
            MAX_STEPS=50,
            GRADIENT_ACCUMULATION_STEPS=1,
            TEST_PER_GPU_BATCH_SIZE=2,
            BEAM_SIZE=3,
            MODEL_DIR=tmp,
            RESULT_DIR=tmp,
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["rouge_1_f_score"], 0.2, abs=ABS_TOL)
    assert pytest.approx(result["rouge_2_f_score"], 0.07, abs=ABS_TOL)
    assert pytest.approx(result["rouge_l_f_score"], 0.16, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_entailment_multinli_bert(notebooks, tmp):
    notebook_path = notebooks["entailment_multinli_transformers"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        parameters={
            "MODEL_NAME": "bert-base-uncased",
            "TO_LOWER": True,
            "TRAIN_DATA_USED_FRACTION": 0.05,
            "DEV_DATA_USED_FRACTION": 0.05,
            "NUM_EPOCHS": 1,
            "CACHE_DIR": tmp
        },
        kernel_name=KERNEL_NAME,
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["matched_precision"], 0.76, abs=ABS_TOL)
    assert pytest.approx(result["matched_recall"], 0.76, abs=ABS_TOL)
    assert pytest.approx(result["matched_f1"], 0.76, abs=ABS_TOL)
    assert pytest.approx(result["mismatched_precision"], 0.76, abs=ABS_TOL)
    assert pytest.approx(result["mismatched_recall"], 0.76, abs=ABS_TOL)
    assert pytest.approx(result["mismatched_f1"], 0.76, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_tc_mnli_transformers(notebooks, tmp):
    notebook_path = notebooks["tc_mnli_transformers"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            NUM_GPUS=1,
            DATA_FOLDER=tmp,
            CACHE_DIR=tmp,
            BATCH_SIZE=16,
            NUM_EPOCHS=1,
            TRAIN_DATA_FRACTION=0.05,
            TEST_DATA_FRACTION=0.05,
            MODEL_NAMES=["distilbert-base-uncased"],
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["accuracy"], 0.885, abs=ABS_TOL)
    assert pytest.approx(result["f1"], 0.885, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_minilm_abstractive_summarization(notebooks, tmp):
    notebook_path = notebooks["minilm_abstractive_summarization"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            QUICK_RUN=True,
            NUM_GPUS=torch.cuda.device_count(),
            TOP_N=100,
            WARMUP_STEPS=5,
            MAX_STEPS=50,
            GRADIENT_ACCUMULATION_STEPS=1,
            TEST_PER_GPU_BATCH_SIZE=2,
            BEAM_SIZE=3,
            CLEANUP_RESULTS=True,
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["rouge_1_f_score"], 0.2, abs=ABS_TOL)
    assert pytest.approx(result["rouge_2_f_score"], 0.07, abs=ABS_TOL)
    assert pytest.approx(result["rouge_l_f_score"], 0.16, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_question_answering_squad_transformers(notebooks, tmp):
    notebook_path = notebooks["question_answering_squad_transformers"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        parameters={
            "TRAIN_DATA_USED_PERCENT": 0.15,
            "DEV_DATA_USED_PERCENT": 0.15,
            "NUM_EPOCHS": 1,
            "MAX_SEQ_LENGTH": 384,
            "DOC_STRIDE": 128,
            "PER_GPU_BATCH_SIZE": 4,
            "MODEL_NAME": "distilbert-base-uncased",
            "DO_LOWER_CASE": True,
            "CACHE_DIR": tmp
        },
        kernel_name=KERNEL_NAME,
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["exact"], 0.55, abs=ABS_TOL)
    assert pytest.approx(result["f1"], 0.70, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_bidaf_deep_dive(
    notebooks, subscription_id, resource_group, workspace_name, workspace_region
):
    notebook_path = notebooks["bidaf_deep_dive"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        parameters={
            "NUM_EPOCHS": 1,
            "config_path": None,
            "PROJECT_FOLDER": "examples/question_answering/bidaf-question-answering",
            "SQUAD_FOLDER": "examples/question_answering/squad",
            "LOGS_FOLDER": "examples/question_answering/",
            "BIDAF_CONFIG_PATH": "examples/question_answering/",
            "subscription_id": subscription_id,
            "resource_group": resource_group,
            "workspace_name": workspace_name,
            "workspace_region": workspace_region,
        },
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict["validation_EM"]
    assert result == pytest.approx(0.5, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_extractive_summarization_cnndm_transformers(notebooks, tmp):
    notebook_path = notebooks["extractive_summarization_cnndm_transformer"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            QUICK_RUN=True,
            TOP_N=100,
            CHUNK_SIZE=200,
            USE_PREPROCESSED_DATA=False,
            DATA_PATH=tmp,
            CACHE_DIR=tmp,
            BATCH_SIZE=3000,
            REPORT_EVERY=50,
            MAX_STEPS=100,
            WARMUP_STEPS=5e2,
            MODEL_NAME="distilbert-base-uncased",
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["rouge_2_f_score"], 0.1, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_extractive_summarization_cnndm_transformers_processed(notebooks, tmp):
    notebook_path = notebooks["extractive_summarization_cnndm_transformer"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            QUICK_RUN=True,
            TOP_N=100,
            CHUNK_SIZE=200,
            USE_PREPROCESSED_DATA=True,
            DATA_PATH=tmp,
            CACHE_DIR=tmp,
            PROCESSED_DATA_PATH=tmp,
            BATCH_SIZE=3000,
            REPORT_EVERY=50,
            MAX_STEPS=100,
            WARMUP_STEPS=5e2,
            MODEL_NAME="distilbert-base-uncased",
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["rouge_2_f_score"], 0.1, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_abstractive_summarization_bertsumabs_cnndm(notebooks, tmp):
    notebook_path = notebooks["abstractive_summarization_bertsumabs_cnndm"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            QUICK_RUN=True,
            TOP_N=1000,
            MAX_POS=512,
            DATA_FOLDER=tmp,
            CACHE_DIR=tmp,
            BATCH_SIZE_PER_GPU=3,
            REPORT_EVERY=50,
            MAX_STEPS=100,
            MODEL_NAME="bert-base-uncased",
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert pytest.approx(result["rouge_2_f_score"], 0.01, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_bert_senteval(
    notebooks, subscription_id, resource_group, workspace_name, workspace_region, tmp
):
    notebook_path = notebooks["bert_senteval"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        kernel_name=KERNEL_NAME,
        parameters=dict(
            subscription_id=subscription_id,
            resource_group=resource_group,
            workspace_name=workspace_name,
            workspace_region=workspace_region,
            CACHE_DIR=tmp,
            LOCAL_UTILS="utils_nlp",
            LOCAL_SENTEVAL="utils_nlp/eval/SentEval",
            EXPERIMENT_NAME="test-nlp-ss-bert",
            CLUSTER_NAME="eval-gpu",
            MAX_NODES=1,
        ),
    )
    pearson = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict["pearson"]
    mse = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict["mse"]
    assert pearson == pytest.approx(0.6, abs=ABS_TOL)
    assert mse < 1.8 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_automl_local_deployment_aci(
    notebooks, subscription_id, resource_group, workspace_name, workspace_region
):
    notebook_path = notebooks["automl_local_deployment_aci"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        parameters={
            "automl_iterations": 1,
            "automl_iteration_timeout": 7,
            "config_path": None,
            "webservice_name": "aci-test-service",
            "subscription_id": subscription_id,
            "resource_group": resource_group,
            "workspace_name": workspace_name,
            "workspace_region": workspace_region,
        },
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict["pearson_correlation"]
    assert result == pytest.approx(0.5, abs=ABS_TOL) 

# 需要导入模块: import scrapbook [as 别名]
# 或者: from scrapbook import read_notebook [as 别名]
def test_gensen_aml_deep_dive(notebooks):
    notebook_path = notebooks["gensen_aml_deep_dive"]
    pm.execute_notebook(
        notebook_path,
        OUTPUT_NOTEBOOK,
        parameters=dict(
            CACHE_DIR="./tests/integration/temp",
            AZUREML_CONFIG_PATH="./tests/integration/.azureml",
            UTIL_NLP_PATH="./utils_nlp",
            MAX_EPOCH=1,
            TRAIN_SCRIPT="./examples/sentence_similarity/gensen_train.py",
            CONFIG_PATH="./examples/sentence_similarity/gensen_config.json",
            MAX_TOTAL_RUNS=1,
            MAX_CONCURRENT_RUNS=1,
        ),
    )
    result = sb.read_notebook(OUTPUT_NOTEBOOK).scraps.data_dict
    assert result["min_val_loss"] > 5
    assert result["learning_rate"] >= 0.0001
    assert result["learning_rate"] <= 0.001