### Redirect example configuration to a file

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Redirect the output of the --example flag to a file to save the example configuration. This example shows saving a training configuration.

```bash
kiwi train --example > train.yaml
```

--------------------------------

### Run Kiwi CLI command with example configuration

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Use the --example flag to print example configuration files to the terminal. This is useful for creating new configuration files.

```bash
kiwi (train|pretrain|predict|evaluate|search) --example
```

--------------------------------

### Run OpenKiwi Help Command

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

After installation, you can verify the installation and view available commands by running the help command.

```bash
python kiwi -h
```

```bash
kiwi -h
```

--------------------------------

### Install OpenKiwi as a Library

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

Use this command to install OpenKiwi as a Python library for use in your projects.

```bash
pip install openkiwi
```

--------------------------------

### Install OpenKiwi as a Local Package with Poetry

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

Installs all dependencies for OpenKiwi when developing locally using Poetry. Ensure Poetry is installed first.

```bash
poetry install
```

--------------------------------

### Install Poetry using Curl

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

Installs the Poetry dependency manager using a recommended script. This is a prerequisite for installing OpenKiwi as a local package.

```bash
curl -sSL https://raw.githubusercontent.com/sdispater/poetry/master/get-poetry.py | python
```

--------------------------------

### Install OpenKiwi with Optuna Search (Poetry)

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

Installs OpenKiwi with optional Optuna integration for hyperparameter search using Poetry. This is useful for optimizing model parameters.

```bash
poetry install -E search
```

--------------------------------

### Configuration Composition Example

Source: https://github.com/unbabel/openkiwi/blob/master/docs/configuration/index.rst

Use a different dataset configuration by specifying it in the command line, leveraging Hydra/OmegaConf's composition capabilities.

```bash
kiwi train config/bert.yaml data=unbabel.qe.en_pt
```

--------------------------------

### Run OpenKiwi from Command Line

Source: https://github.com/unbabel/openkiwi/blob/master/README.md

Execute the OpenKiwi command-line interface after installation.

```bash
kiwi
```

--------------------------------

### Install OpenKiwi with MLflow Integration (Poetry)

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

Installs OpenKiwi with optional MLflow integration using Poetry. This enables advanced ML tracking capabilities.

```bash
poetry install -E mlflow
```

--------------------------------

### Example BERT Search Configuration

Source: https://github.com/unbabel/openkiwi/blob/master/docs/configuration/search.rst

This is an example search configuration file for the BERT model. It should be placed in the `config/` directory.

```yaml
search:
  model: BERT
  # ... other BERT specific parameters

```

--------------------------------

### Install OpenKiwi with Optuna Search (Pip)

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

Installs OpenKiwi with optional Optuna integration for hyperparameter search using pip. This is useful for optimizing model parameters.

```bash
pip install openkiwi[search]
```

--------------------------------

### Install OpenKiwi with MLflow Integration (Pip)

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

Installs OpenKiwi with optional MLflow integration using pip. This enables advanced ML tracking capabilities.

```bash
pip install openkiwi[mlflow]
```

--------------------------------

### Install Poetry using Pip

Source: https://github.com/unbabel/openkiwi/blob/master/docs/installation.rst

An alternative method to install Poetry using pip. Note that this is not recommended as it may conflict with local dependencies.

```bash
pip install poetry
```

--------------------------------

### YAML Configuration with Defaults

Source: https://github.com/unbabel/openkiwi/blob/master/docs/configuration/index.rst

Example of a YAML configuration file that uses defaults to load another configuration file, demonstrating configuration composition.

```yaml
defaults:
 - data: wmt19.qe.en_de
```

--------------------------------

### CLI Override Example

Source: https://github.com/unbabel/openkiwi/blob/master/docs/configuration/index.rst

Override configuration values when running a Kiwi command using dot notation for nested keys.

```bash
kiwi train config/bert.yaml trainer.gpus=0 system.batch_size=16
```

--------------------------------

### Import OpenKiwi in Python

Source: https://github.com/unbabel/openkiwi/blob/master/README.md

Import the kiwi module into your Python project after installation.

```python
import kiwi
```

--------------------------------

### Pretrain Predictor model using CLI

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

This command initiates the pretraining process for the Predictor model using a specified configuration file.

```bash
kiwi pretrain config/predictor.yaml
```

--------------------------------

### Python Search from Configuration Dictionary

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Perform a search using a configuration dictionary loaded from a file. This method involves first converting the file to a dictionary, then passing it to the search function. Requires importing both file_to_configuration and search_from_configuration.

```python
from kiwi.lib.search import search_from_configuration
from kiwi.lib.utils import file_to_configuration

configuration_dict = file_to_configuration('config/search.yaml')
optuna_study = search_from_configuration(configuration_dict)
```

--------------------------------

### Train BERT model for QE using Python API (dictionary config)

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Load a configuration from a file into a dictionary and then use train_from_configuration to train a BERT model for QE. This method allows for other file formats like JSON.

```python
from kiwi.lib.train import train_from_configuration
from kiwi.lib.utils import file_to_configuration

configuration_dict = file_to_configuration('config/bert.yaml')
run_info = train_from_configuration(configuration_dict)
```

--------------------------------

### Train BERT model for QE using CLI

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

This command trains a BERT model for Quality Estimation (QE) using a specified configuration file.

```bash
kiwi train config/bert.yaml
```

--------------------------------

### Train BERT model for QE using Python API (file config)

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Use the train_from_file function from kiwi.lib.train to train a BERT model for QE, providing the configuration file path.

```python
from kiwi.lib.train import train_from_file

run_info = train_from_file('config/bert.yaml')
```

--------------------------------

### Command-line Search

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Execute a search directly from the command line using a YAML configuration file.

```bash
kiwi search config/search.yaml
```

--------------------------------

### Predict on a dataset using Python API

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Load a configuration file and use predict_from_configuration to generate predictions and metrics for a full dataset. The configuration specifies model and data paths.

```python
from kiwi.lib.predict import predict_from_configuration
from kiwi.lib.utils import file_to_configuration

configuration_dict = file_to_configuration('config/predict.yaml')
predictions, metrics = predict_from_configuration(configuration_dict)
```

--------------------------------

### Predict using a trained QE model via CLI

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Use the predict pipeline via the CLI by providing a YAML configuration file that specifies model path, data locations, and output directories.

```bash
kiwi predict config/predict.yaml
```

--------------------------------

### Python Search from File

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Perform a search by loading the configuration directly from a YAML file using the Python API. Requires importing the search_from_file function.

```python
from kiwi.lib.search import search_from_file

optuna_study = search_from_file('config/search.yaml')
```

--------------------------------

### Evaluation Configuration YAML

Source: https://github.com/unbabel/openkiwi/blob/master/docs/configuration/evaluate.rst

This snippet shows the structure of the evaluation configuration file. Use this to define parameters for evaluation runs.

```yaml
model:
  name: "bert-base-uncased"
  path: "/path/to/your/model"

data:
  name: "sst2"
  path: "/path/to/your/data"

output:
  path: "/path/to/your/output"

metrics:
  - "accuracy"
  - "f1"

params:
  batch_size: 32
  epochs: 3
```

--------------------------------

### Evaluate model predictions using Python API

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Use evaluate_from_configuration to programmatically evaluate predictions against a gold standard file. The configuration dictionary should specify prediction and reference file details.

```python
from kiwi.lib.evaluate import evaluate_from_configuration
from kiwi.lib.utils import file_to_configuration

configuration_dict = file_to_configuration('config/evaluate.yaml')
report = evaluate_from_configuration(configuration_dict)
print(report)
```

--------------------------------

### Evaluate model predictions using CLI

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Evaluate the performance of a trained model's predictions against a reference file using the CLI. A YAML configuration file specifies the formats and locations of predictions and reference data.

```bash
kiwi evaluate config/predict.yaml
```

--------------------------------

### Bert Model Training Configuration

Source: https://github.com/unbabel/openkiwi/blob/master/docs/configuration/train.rst

YAML configuration file for the Bert model training. This file specifies various parameters for training the Bert model.

```yaml
model:
  name: bert
  pretrained: bert-base-uncased
  encoder_layers: 6
  decoder_layers: 6
  d_model: 512
  num_heads: 8
  d_ff: 2048
  dropout: 0.1
  label_smoothing: 0.1

data:
  train_path: data/wmt14-ende/train.txt
  val_path: data/wmt14-ende/valid.txt
  test_path: data/wmt14-ende/test.txt
  batch_size: 32
  max_seq_len: 512
  num_workers: 4

training:
  optimizer: adam
  learning_rate: 0.0001
  warmup_steps: 4000
  max_steps: 100000
  log_steps: 100
  save_steps: 5000
  eval_steps: 5000
  gradient_accumulation_steps: 1
  fp16: true

logging:
  level: info
  file: logs/bert.log

checkpoint:
  path: checkpoints/bert
  resume: false

```

--------------------------------

### Load a trained model for on-demand predictions

Source: https://github.com/unbabel/openkiwi/blob/master/docs/usage.rst

Use load_system to load a trained model checkpoint and keep it in memory for on-demand predictions on source and target text pairs.

```python
from kiwi.lib.predict import load_system

runner = load_system('trained_models/model.ckpt')
predictions = runner.predict(
    source=['Aqui vai um exemplo de texto'],
    target=['Here is an example text'],
)
```

--------------------------------

### Prediction Configuration File

Source: https://github.com/unbabel/openkiwi/blob/master/docs/configuration/predict.rst

This snippet shows the structure of the prediction configuration file. It is used to define parameters for prediction tasks.

```yaml
model:
  name: "bert-base-uncased"
  path: "/opt/ml/model"
  type: "transformer"
  device: "cuda"

input:
  max_length: 128
  batch_size: 32

output:
  path: "/opt/ml/output"
  format: "json"

logging:
  level: "INFO"
  file: "/opt/ml/logs/predict.log"

```

--------------------------------

### OpenKiwi Citation

Source: https://github.com/unbabel/openkiwi/blob/master/README.md

BibTeX entry for citing the OpenKiwi paper. Use this in academic publications.

```bibtex
@inproceedings{openkiwi,
    author = {Fábio Kepler and
              Jonay Trénous and
              Marcos Treviso and
              Miguel Vera and
              André F. T. Martins},
    title  = {Open{K}iwi: An Open Source Framework for Quality Estimation},
    year   = {2019},
    booktitle = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics--System Demonstrations},
    pages  = {117--122},
    month  = {July},
    address = {Florence, Italy},
    url    = {https://www.aclweb.org/anthology/P19-3020},
    organization = {Association for Computational Linguistics},
}
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.