Merge remote-tracking branch 'origin/main' into whoisjones/improved_n…

…aming_for_sampling

# Conflicts:
#	tests/test_dataset_generator.py

README.md

@@ -1,4 +1,6 @@
-<h1 align="center">Dataset Generator</h1>
+![Fabricator Logo](resources/logo_fabricator.drawio_dark.png#gh-dark-mode-only)
+![Fabricator Logo](resources/logo_fabricator.drawio_white.png#gh-light-mode-only)
+
 <p align="center">A flexible open-source framework to generate datasets with large language models.</p>
 <p align="center">
 <img alt="version" src="https://img.shields.io/badge/version-0.1-green">
@@ -8,8 +10,8 @@
 <div align="center">
 <hr>
 
-[Installation](#installation) - [Basic Concepts](#basic-concepts) - [Examples](#examples) - [Tutorials](tutorials/TUTORIAL-1_OVERVIEW.md) - 
-Paper - [Citation](#citation)
+[Installation](#installation) | [Basic Concepts](#basic-concepts) | [Examples](#examples) | [Tutorials](tutorials/TUTORIAL-1_OVERVIEW.md) | 
+Paper | [Citation](#citation)
 
 <hr>
 </div>
@@ -30,10 +32,10 @@ prompt customization, integration and sampling of fewshot examples or annotation
 ## Installation
 Using conda:
 ```
-git clone [email protected]:flairNLP/ai-dataset-generator.git
-cd ai-dataset-generator
-conda create -y -n aidatasetgenerator python=3.10
-conda activate aidatasetgenerator
+git clone [email protected]:flairNLP/fabricator.git
+cd fabricator
+conda create -y -n fabricator python=3.10
+conda activate fabricator
 pip install -e .
 ```
 
@@ -45,10 +47,9 @@ we need four basic modules: a dataset, a prompt, a language model and a generato
 unlabeled datasets and store the generated or annotated datasets with their `Dataset` class. Once
 created, you can share the dataset with others via the hub or use it for your model training.
 - <b>Prompt</b>: A prompt is the instruction made to the language model. It can be a simple sentence or a more complex
-template with placeholders. We utilize [langchain](https://github.com/langchain-ai/langchain) `PromptTemplate` classes
-and provide an easy interface for custom dataset generation prompts in which you can specify label options
-for the LLM to choose from, provide fewshot examples to support the prompt with or annotate an unlabeled dataset
-in a specific way.
+template with placeholders. We provide an easy interface for custom dataset generation prompts in which you can specify 
+label options for the LLM to choose from, provide fewshot examples to support the prompt with or annotate an unlabeled 
+dataset in a specific way.
 - <b>LLM</b>: We use [deepset's haystack library](https://github.com/deepset-ai/haystack) as our LLM interface. deepset
 supports a wide range of LLMs including OpenAI, all models from the HuggingFace model hub and many more.
 - <b>Generator</b>: The generator is the core of this framework. It takes a dataset, a prompt and a LLM and generates a
@@ -64,8 +65,8 @@ as that:
 ```python
 import os
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator import DatasetGenerator
+from fabricator.prompts import BasePrompt
 
 prompt = BasePrompt(
     task_description="Generate a short movie review.",

paper_experiments/conll_annotate_dataset.py

@@ -1,8 +1,8 @@
 import os
 from datasets import load_dataset
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator, BasePrompt
-from ai_dataset_generator.dataset_transformations.token_classification import convert_token_labels_to_spans
+from fabricator import DatasetGenerator, BasePrompt
+from fabricator.dataset_transformations.token_classification import convert_token_labels_to_spans
 
 
 def run():

paper_experiments/conll_gpt_train_model.py

@@ -1,6 +1,6 @@
 from argparse import ArgumentParser
 from datasets import load_dataset
-from ai_dataset_generator import convert_spans_to_token_labels
+from fabricator import convert_spans_to_token_labels
 from seqeval.metrics import accuracy_score, f1_score
 
 

paper_experiments/mrpc_annotate_dataset.py

@@ -1,8 +1,8 @@
 import os
 from datasets import load_dataset, concatenate_datasets
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator, BasePrompt
-from ai_dataset_generator.dataset_transformations.text_classification import convert_label_ids_to_texts
+from fabricator import DatasetGenerator, BasePrompt
+from fabricator.dataset_transformations.text_classification import convert_label_ids_to_texts
 
 
 def run():

paper_experiments/snli_annotate_dataset.py

@@ -1,8 +1,8 @@
 import os
 from datasets import load_dataset, concatenate_datasets
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator, BasePrompt
-from ai_dataset_generator.dataset_transformations.text_classification import convert_label_ids_to_texts
+from fabricator import DatasetGenerator, BasePrompt
+from fabricator.dataset_transformations.text_classification import convert_label_ids_to_texts
 
 
 def run():

paper_experiments/squad_annotate_dataset.py

@@ -4,8 +4,8 @@
 
 from datasets import Sequence, Value, load_dataset, concatenate_datasets
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator, BasePrompt
-from ai_dataset_generator.dataset_transformations.question_answering import (
+from fabricator import DatasetGenerator, BasePrompt
+from fabricator.dataset_transformations.question_answering import (
     preprocess_squad_format,
     postprocess_squad_format,
 )

paper_experiments/trec_annotate_dataset.py

@@ -1,8 +1,8 @@
 import os
 from datasets import load_dataset, concatenate_datasets
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator, BasePrompt
-from ai_dataset_generator.dataset_transformations.text_classification import convert_label_ids_to_texts
+from fabricator import DatasetGenerator, BasePrompt
+from fabricator.dataset_transformations.text_classification import convert_label_ids_to_texts
 
 
 def run():

paper_experiments/trec_generate_dataset.py

@@ -1,8 +1,8 @@
 import os
 from datasets import load_dataset, concatenate_datasets
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator, BasePrompt
-from ai_dataset_generator.dataset_transformations.text_classification import convert_label_ids_to_texts
+from fabricator import DatasetGenerator, BasePrompt
+from fabricator.dataset_transformations.text_classification import convert_label_ids_to_texts
 
 
 def run():

paper_experiments/trec_hyperparameter_generate_dataset.py

@@ -1,8 +1,8 @@
 import os
 from datasets import load_dataset, concatenate_datasets
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator, BasePrompt
-from ai_dataset_generator.dataset_transformations.text_classification import convert_label_ids_to_texts
+from fabricator import DatasetGenerator, BasePrompt
+from fabricator.dataset_transformations.text_classification import convert_label_ids_to_texts
 
 def run():
     for possible_examples_per_class, fewshot_example_per_class in [(0,0), (2,2), (4,2), (4,3), (4,4), (8,2), (8,3),

src/ai_dataset_generator/dataset_generator.py → src/fabricator/dataset_generator.py

@@ -18,7 +18,7 @@
 
 
 class DatasetGenerator:
-    """The DatasetGenerator class is the main class of the ai_dataset_generator package.
+    """The DatasetGenerator class is the main class of the fabricator package.
     It generates datasets based on a prompt template. The main function is generate()."""
 
     def __init__(self, prompt_node: PromptNode, max_tries: int = 10):

tests/test_dataset_generator.py

@@ -1,9 +1,10 @@
 import unittest
 
 from datasets import Dataset, load_dataset
-from ai_dataset_generator import DatasetGenerator
-from ai_dataset_generator.prompts import BasePrompt
-from ai_dataset_generator.dataset_transformations.text_classification import convert_label_ids_to_texts
+
+from fabricator import DatasetGenerator
+from fabricator.prompts import BasePrompt
+from fabricator.dataset_transformations.text_classification import convert_label_ids_to_texts
 
 
 class TestDatasetGenerator(unittest.TestCase):

tests/test_dataset_sampler.py

@@ -3,7 +3,7 @@
 from collections import Counter
 from datasets import load_dataset
 
-from ai_dataset_generator.samplers import random_sampler, single_label_task_sampler, ml_mc_sampler, \
+from fabricator.samplers import random_sampler, single_label_task_sampler, ml_mc_sampler, \
     single_label_stratified_sample
 
 

tests/test_dataset_transformations.py

@@ -2,10 +2,10 @@
 
 from datasets import load_dataset
 
-from ai_dataset_generator.prompts import BasePrompt
-from ai_dataset_generator.dataset_transformations.question_answering import *
-from ai_dataset_generator.dataset_transformations.text_classification import *
-from ai_dataset_generator.dataset_transformations.token_classification import *
+from fabricator.prompts import BasePrompt
+from fabricator.dataset_transformations.question_answering import *
+from fabricator.dataset_transformations.text_classification import *
+from fabricator.dataset_transformations.token_classification import *
 
 
 class TestTransformationsTextClassification(unittest.TestCase):

tests/test_prompts.py

@@ -2,7 +2,7 @@
 
 from datasets import load_dataset, Dataset, QuestionAnsweringExtractive, TextClassification, Summarization
 
-from ai_dataset_generator.prompts import (
+from fabricator.prompts import (
     BasePrompt,
     infer_prompt_from_task_template,
 )

tutorials/TUTORIAL-1_OVERVIEW.md

@@ -57,7 +57,7 @@ your prompt generates exactly one data point per prompt since the output will be
 ### Create a minimal prompt for generating text without fewshot examples
 
 ```python
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator.prompts import BasePrompt
 
 prompt_template = BasePrompt(task_description="Generate movie reviews.")
 print(prompt_template.get_prompt_text())
@@ -75,7 +75,7 @@ options are a list of strings. When generating data, the `DatasetGenerator` clas
 one of the label options and insert it into the task description such that the generated dataset is balanced.
 
 ```python
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator.prompts import BasePrompt
 
 label_options = ["positive", "negative"]
 prompt_template = BasePrompt(
@@ -109,7 +109,7 @@ fewshot dataset that have the same label as used in the task description as exem
 
 ```python
 from datasets import Dataset
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator.prompts import BasePrompt
 
 label_options = ["positive", "negative"]
 
@@ -152,7 +152,7 @@ annotated by the LLM as illustrated in the previous example.
 
 ```python
 from datasets import Dataset
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator.prompts import BasePrompt
 
 label_options = ["positive", "negative"]
 
@@ -208,7 +208,7 @@ frameworks such as `transformers`.
 
 ```python
 from datasets import Dataset
-from ai_dataset_generator import DatasetGenerator
+from fabricator import DatasetGenerator
 
 fewshot_examples = Dataset.from_dict({
     "text": ["This movie is great!", "This movie is bad!"],

tutorials/TUTORIAL-2_SIMPLE-GENERATION.md

@@ -9,8 +9,8 @@ to the HuggingFace Hub.
 ```python
 import os
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator import DatasetGenerator
+from fabricator.prompts import BasePrompt
 
 prompt = BasePrompt(
     task_description="Generate a very very short movie review.",
@@ -39,8 +39,8 @@ this can be achieved by providing a `label_options` argument to the `BasePrompt`
 ```python
 import os
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator import DatasetGenerator
+from fabricator.prompts import BasePrompt
 
 label_options = ["positive", "negative"]
 
@@ -87,8 +87,8 @@ sampling. In this case, we use the `label` column.
 import os
 from datasets import Dataset
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator import DatasetGenerator
+from fabricator.prompts import BasePrompt
 
 label_options = ["positive", "negative"]
 
@@ -138,8 +138,8 @@ argument since the generator will use the column specified in `generate_data_for
 import os
 from datasets import Dataset
 from haystack.nodes import PromptNode
-from ai_dataset_generator import DatasetGenerator
-from ai_dataset_generator.prompts import BasePrompt
+from fabricator import DatasetGenerator
+from fabricator.prompts import BasePrompt
 
 label_options = ["positive", "negative"]