Axolotl

# Axolotl - Api

**Pages:** 150

---

## cli.cloud.modal_

**URL:** https://docs.axolotl.ai/docs/api/cli.cloud.modal_.html

**Contents:**
- cli.cloud.modal_
- Classes
  - ModalCloud
- Functions
  - run_cmd

Modal Cloud support from CLI

Modal Cloud implementation.

Run a command inside a folder, with Modal Volume reloading before and commit on success.

**Examples:**

Example 1 (python):
```python
cli.cloud.modal_.ModalCloud(config, app=None)
```

Example 2 (python):
```python
cli.cloud.modal_.run_cmd(cmd, run_folder, volumes=None)
```

---

## core.trainers.base

**URL:** https://docs.axolotl.ai/docs/api/core.trainers.base.html

**Contents:**
- core.trainers.base
- Classes
  - AxolotlTrainer
    - Methods
      - log
        - Parameters
      - push_to_hub
      - store_metrics
        - Parameters

Module for customized trainers

Extend the base Trainer for axolotl helpers

Log logs on the various objects watching training, including stored metrics.

Overwrite the push_to_hub method in order to force-add the tags when pushing the model on the Hub. Please refer to ~transformers.Trainer.push_to_hub for more details.

Store metrics with specified reduction type.

**Examples:**

Example 1 (python):
```python
core.trainers.base.AxolotlTrainer(
    *_args,
    bench_data_collator=None,
    eval_data_collator=None,
    dataset_tags=None,
    **kwargs,
)
```

Example 2 (python):
```python
core.trainers.base.AxolotlTrainer.log(logs, start_time=None)
```

Example 3 (python):
```python
core.trainers.base.AxolotlTrainer.push_to_hub(*args, **kwargs)
```

Example 4 (python):
```python
core.trainers.base.AxolotlTrainer.store_metrics(
    metrics,
    train_eval='train',
    reduction='mean',
)
```

---

## prompt_strategies.input_output

**URL:** https://docs.axolotl.ai/docs/api/prompt_strategies.input_output.html

**Contents:**
- prompt_strategies.input_output
- Classes
  - RawInputOutputPrompter
  - RawInputOutputStrategy

prompt_strategies.input_output

Module for plain input/output prompt pairs

prompter for raw i/o data

Prompt Strategy class for input/output pairs

**Examples:**

Example 1 (python):
```python
prompt_strategies.input_output.RawInputOutputPrompter()
```

Example 2 (python):
```python
prompt_strategies.input_output.RawInputOutputStrategy(
    *args,
    eos_token=None,
    **kwargs,
)
```

---

## prompt_strategies.completion

**URL:** https://docs.axolotl.ai/docs/api/prompt_strategies.completion.html

**Contents:**
- prompt_strategies.completion
- Classes
  - CompletionPromptTokenizingStrategy
  - CompletionPrompter

prompt_strategies.completion

Basic completion text

Tokenizing strategy for Completion prompts.

Prompter for completion

**Examples:**

Example 1 (python):
```python
prompt_strategies.completion.CompletionPromptTokenizingStrategy(
    *args,
    max_length=None,
    **kwargs,
)
```

Example 2 (python):
```python
prompt_strategies.completion.CompletionPrompter()
```

---

## utils.collators.core

**URL:** https://docs.axolotl.ai/docs/api/utils.collators.core.html

**Contents:**
- utils.collators.core

basic shared collator constants

---

## monkeypatch.data.batch_dataset_fetcher

**URL:** https://docs.axolotl.ai/docs/api/monkeypatch.data.batch_dataset_fetcher.html

**Contents:**
- monkeypatch.data.batch_dataset_fetcher
- Functions
  - apply_multipack_dataloader_patch
  - patch_fetchers
  - patched_worker_loop
  - remove_multipack_dataloader_patch

monkeypatch.data.batch_dataset_fetcher

Monkey patches for the dataset fetcher to handle batches of packed indexes.

This patch allows DataLoader to correctly process batches that contain multiple bins of packed sequences.

Apply patches to PyTorch’s DataLoader components.

Worker loop that ensures patches are applied in worker processes.

Remove the monkeypatch and restore original PyTorch DataLoader behavior.

**Examples:**

Example 1 (python):
```python
monkeypatch.data.batch_dataset_fetcher.apply_multipack_dataloader_patch()
```

Example 2 (python):
```pyt