Code Monkey home page Code Monkey logo

Comments (1)

sweep-ai avatar sweep-ai commented on July 22, 2024

🚀 Here's the PR! #50

See Sweep's progress at the progress dashboard!
💎 Sweep Pro: I'm using GPT-4. You have unlimited GPT-4 tickets. (tracking ID: dc339794ff)

Actions (click)

  • ↻ Restart Sweep

Sandbox execution failed

The sandbox appears to be unavailable or down.

Step 1: 🔎 Searching

I found the following snippets in your repository. I will now analyze these snippets and come up with a plan.

Some code snippets I think are relevant in decreasing order of relevance (click to expand). If some file is missing from here, you can mention the path in the ticket description.

dspy/dspy/datasets/dataset.py

Lines 6 to 112 in c7a522a

class Dataset:
def __init__(self, train_seed=0, train_size=None, eval_seed=0, dev_size=None, test_size=None):
self.train_size = train_size
self.train_seed = train_seed
self.dev_size = dev_size
self.dev_seed = eval_seed
self.test_size = test_size
self.test_seed = eval_seed
self.do_shuffle = True
self.name = self.__class__.__name__
def reset_seeds(self, train_seed=None, train_size=None, eval_seed=None, dev_size=None, test_size=None):
self.train_size = train_size if train_size is not None else self.train_size
self.train_seed = train_seed if train_seed is not None else self.train_seed
self.dev_size = dev_size if dev_size is not None else self.dev_size
self.dev_seed = eval_seed if eval_seed is not None else self.dev_seed
self.test_size = test_size if test_size is not None else self.test_size
self.test_seed = eval_seed if eval_seed is not None else self.test_seed
if hasattr(self, '_train_'):
del self._train_
if hasattr(self, '_dev_'):
del self._dev_
if hasattr(self, '_test_'):
del self._test_
@property
def train(self):
if not hasattr(self, '_train_'):
self._train_ = self._shuffle_and_sample('train', self._train, self.train_size, self.train_seed)
return self._train_
@property
def dev(self):
if not hasattr(self, '_dev_'):
self._dev_ = self._shuffle_and_sample('dev', self._dev, self.dev_size, self.dev_seed)
return self._dev_
@property
def test(self):
if not hasattr(self, '_test_'):
self._test_ = self._shuffle_and_sample('test', self._test, self.test_size, self.test_seed)
return self._test_
def _shuffle_and_sample(self, split, data, size, seed=0):
'''
The setting (seed=s, size=N) is always a subset
of the setting (seed=s, size=M) for N < M.
'''
data = list(data)
# Shuffle the data irrespective of the requested size.
base_rng = random.Random(seed)
if self.do_shuffle:
base_rng.shuffle(data)
data = data[:size]
output = []
for example in data:
output.append(Example(**example, dspy_uuid=str(uuid.uuid4()), dspy_split=split))
# TODO: NOTE: Ideally we use these uuids for dedup internally, for demos and internal train/val splits.
# Now, some tasks (like convQA and Colors) have overlapping examples. Here, we should allow the user to give us
# a uuid field that would respect this in some way. This means that we need a more refined concept that
# uuid (each example is unique) and more like a group_uuid.
# rng = random.Random(seed)
# rng.shuffle(data)
return output
@classmethod
def prepare_by_seed(cls, train_seeds=[1,2,3,4,5], train_size=16, dev_size=1000,
divide_eval_per_seed=True, eval_seed=2023, **kwargs):
data_args = dotdict(train_size=train_size, eval_seed=eval_seed, dev_size=dev_size, test_size=0, **kwargs)
dataset = cls(**data_args)
eval_set = dataset.dev
eval_sets, train_sets = [], []
examples_per_seed = dev_size // len(train_seeds) if divide_eval_per_seed else dev_size
eval_offset = 0
for train_seed in train_seeds:
data_args.train_seed = train_seed
dataset.reset_seeds(**data_args)
eval_sets.append(eval_set[eval_offset:eval_offset+examples_per_seed])
train_sets.append(dataset.train)
assert len(eval_sets[-1]) == examples_per_seed, len(eval_sets[-1])
assert len(train_sets[-1]) == train_size, len(train_sets[-1])
if divide_eval_per_seed:
eval_offset += examples_per_seed

dspy/dspy/datasets/gsm8k.py

Lines 4 to 84 in c7a522a

from datasets import load_dataset
from dspy.datasets.dataset import Dataset
class GSM8K:
def __init__(self) -> None:
super().__init__()
self.do_shuffle = False
dataset = load_dataset("gsm8k", 'main')
hf_official_train = dataset['train']
hf_official_test = dataset['test']
official_train = []
official_test = []
for example in tqdm.tqdm(hf_official_train):
question = example['question']
answer = example['answer'].strip().split()
assert answer[-2] == '####'
gold_reasoning = ' '.join(answer[:-2])
answer = str(int(answer[-1].replace(',', '')))
official_train.append(dict(question=question, gold_reasoning=gold_reasoning, answer=answer))
for example in tqdm.tqdm(hf_official_test):
question = example['question']
answer = example['answer'].strip().split()
assert answer[-2] == '####'
gold_reasoning = ' '.join(answer[:-2])
answer = str(int(answer[-1].replace(',', '')))
official_test.append(dict(question=question, gold_reasoning=gold_reasoning, answer=answer))
rng = random.Random(0)
rng.shuffle(official_train)
rng = random.Random(0)
rng.shuffle(official_test)
trainset = official_train[:200]
devset = official_train[200:500]
testset = official_test[:]
import dspy
trainset = [dspy.Example(**x).with_inputs('question') for x in trainset]
devset = [dspy.Example(**x).with_inputs('question') for x in devset]
testset = [dspy.Example(**x).with_inputs('question') for x in testset]
# print(f"Trainset size: {len(trainset)}")
# print(f"Devset size: {len(devset)}")
# print(f"Testset size: {len(testset)}")
self.train = trainset
self.dev = devset
self.test = testset
def parse_integer_answer(answer, only_first_line=True):
try:
if only_first_line:
answer = answer.strip().split('\n')[0]
# find the last token that has a number in it
answer = [token for token in answer.split() if any(c.isdigit() for c in token)][-1]
answer = answer.split('.')[0]
answer = ''.join([c for c in answer if c.isdigit()])
answer = int(answer)
except (ValueError, IndexError):
# print(answer)
answer = 0
return answer

dspy/dspy/datasets/hotpotqa.py

Lines 6 to 56 in c7a522a

class HotPotQA(Dataset):
def __init__(self, *args, only_hard_examples=True, keep_details='dev_titles', unofficial_dev=True, **kwargs) -> None:
super().__init__(*args, **kwargs)
assert only_hard_examples, "Care must be taken when adding support for easy examples." \
"Dev must be all hard to match official dev, but training can be flexible."
hf_official_train = load_dataset("hotpot_qa", 'fullwiki', split='train')
hf_official_dev = load_dataset("hotpot_qa", 'fullwiki', split='validation')
official_train = []
for raw_example in hf_official_train:
if raw_example['level'] == 'hard':
if keep_details is True:
keys = ['id', 'question', 'answer', 'type', 'supporting_facts']
elif keep_details == 'dev_titles':
keys = ['question', 'answer', 'supporting_facts']
else:
keys = ['question', 'answer']
example = {k: raw_example[k] for k in keys}
if 'supporting_facts' in example:
example['gold_titles'] = set(example['supporting_facts']['title'])
del example['supporting_facts']
official_train.append(example)
rng = random.Random(0)
rng.shuffle(official_train)
self._train = official_train[:len(official_train)*75//100]
if unofficial_dev:
self._dev = official_train[len(official_train)*75//100:]
else:
self._dev = None
for example in self._train:
if keep_details == 'dev_titles':
del example['gold_titles']
test = []
for raw_example in hf_official_dev:
assert raw_example['level'] == 'hard'
example = {k: raw_example[k] for k in ['id', 'question', 'answer', 'type', 'supporting_facts']}
if 'supporting_facts' in example:
example['gold_titles'] = set(example['supporting_facts']['title'])
del example['supporting_facts']
test.append(example)

dspy/dspy/datasets/colors.py

Lines 6 to 30 in c7a522a

class Colors(Dataset):
def __init__(self, sort_by_suffix=True, *args, **kwargs) -> None:
super().__init__(*args, **kwargs)
self.sort_by_suffix = sort_by_suffix
colors = self.sorted_by_suffix(all_colors)
train_size = int(len(colors) * 0.6) # chosen to ensure that similar colors aren't repeated between train and dev
train_colors, dev_colors = colors[:train_size], colors[train_size:]
self._train = [dict(color=color) for color in train_colors]
self._dev = [dict(color=color) for color in dev_colors]
random.Random(0).shuffle(self._train)
random.Random(0).shuffle(self._dev)
def sorted_by_suffix(self, colors):
if not self.sort_by_suffix:
return colors
if isinstance(colors[0], str):
sorted_colors = sorted(colors, key=lambda x: x[::-1])
else:
sorted_colors = sorted(colors, key=lambda x: x['color'][::-1])

dspy/dspy/datasets/__init__.py

Line 1 in c7a522a

from .dataset import Dataset

dspy/README.md

Lines 1 to 386 in c7a522a

current docs link: https://dspy-ai.readthedocs.io/en/latest/
<p align="center">
<img align="center" src="docs/images/DSPy8.png" width="460px" />
</p>
<p align="left">
## DSPy: _Programming_—not prompting—Foundation Models
Paper —— **[DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines](https://arxiv.org/abs/2310.03714)**
[<img align="center" src="https://colab.research.google.com/assets/colab-badge.svg" />](https://colab.research.google.com/github/stanfordnlp/dspy/blob/main/intro.ipynb)
**DSPy** is the framework for solving advanced tasks with language models (LMs) and retrieval models (RMs). **DSPy** unifies techniques for **prompting** and **fine-tuning** LMs — and approaches for **reasoning**, **self-improvement**, and **augmentation with retrieval and tools**. All of these are expressed through modules that compose and learn.
To make this possible:
- **DSPy** provides **composable and declarative modules** for instructing LMs in a familiar Pythonic syntax. It upgrades "prompting techniques" like chain-of-thought and self-reflection from hand-adapted _string manipulation tricks_ into truly modular _generalized operations that learn to adapt to your task_.
- **DSPy** introduces an **automatic compiler that teaches LMs** how to conduct the declarative steps in your program. Specifically, the **DSPy compiler** will internally _trace_ your program and then **craft high-quality prompts for large LMs (or train automatic finetunes for small LMs)** to teach them the steps of your task.
The **DSPy compiler** _bootstraps_ prompts and finetunes from minimal data **without needing manual labels for the intermediate steps** in your program. Instead of brittle "prompt engineering" with hacky string manipulation, you can explore a systematic space of modular and trainable pieces.
For complex tasks, **DSPy** can routinely teach powerful models like `GPT-3.5` and local models like `T5-base` or `Llama2-13b` to be much more reliable at tasks. **DSPy** will compile the _same program_ into different few-shot prompts and/or finetunes for each LM.
If you want to see **DSPy** in action, **[open our intro tutorial notebook](intro.ipynb)**.
### Table of Contents
1. **[Installation](#1-installation)**
1. **[Framework Syntax](#2-syntax-youre-in-charge-of-the-workflowits-free-form-python-code)**
1. **[Compiling: Two Powerful Concepts](#3-two-powerful-concepts-signatures--teleprompters)**
1. **[Tutorials & Documentation](#4-documentation--tutorials)**
1. **[FAQ: Is DSPy right for me?](#5-faq-is-dspy-right-for-me)**
### Analogy to Neural Networks
When we build neural networks, we don't write manual _for-loops_ over lists of _hand-tuned_ floats. Instead, you might use a framework like [PyTorch](https://pytorch.org/) to compose declarative layers (e.g., `Convolution` or `Dropout`) and then use optimizers (e.g., SGD or Adam) to learn the parameters of the network.
Ditto! **DSPy** gives you the right general-purpose modules (e.g., `ChainOfThought`, `Retrieve`, etc.) and takes care of optimizing their prompts _for your program_ and your metric, whatever they aim to do. Whenever you modify your code, your data, or your validation constraints, you can _compile_ your program again and **DSPy** will create new effective prompts that fit your changes.
## 1) Installation
All you need is:
```
pip install dspy-ai
```
Or open our intro notebook in Google Colab: [<img align="center" src="https://colab.research.google.com/assets/colab-badge.svg" />](https://colab.research.google.com/github/stanfordnlp/dspy/blob/main/intro.ipynb)
> _Note: If you're looking for Demonstrate-Search-Predict (DSP), which is the previous version of DSPy, you can find it on the [v1](https://github.com/stanfordnlp/dspy/tree/v1) branch of this repo._
For the optional Pinecone, Qdrant, [chromadb](https://github.com/chroma-core/chroma), or [marqo](https://github.com/marqo-ai/marqo) retrieval integration(s), include the extra(s) below:
```
pip install dspy-ai[pinecone] # or [qdrant] or [chromadb] or [marqo]
```
## 2) Syntax: You're in charge of the workflow—it's free-form Python code!
**DSPy** hides tedious prompt engineering, but it cleanly exposes the important decisions you need to make: **[1]** what's your system design going to look like? **[2]** what are the important constraints on the behavior of your program?
You express your system as free-form Pythonic modules. **DSPy** will tune the quality of your program _in whatever way_ you use foundation models: you can code with loops, `if` statements, or exceptions, and use **DSPy** modules within any Python control flow you think works for your task.
Suppose you want to build a simple retrieval-augmented generation (RAG) system for question answering. You can define your own `RAG` program like this:
```python
class RAG(dspy.Module):
def __init__(self, num_passages=3):
super().__init__()
self.retrieve = dspy.Retrieve(k=num_passages)
self.generate_answer = dspy.ChainOfThought("context, question -> answer")
def forward(self, question):
context = self.retrieve(question).passages
answer = self.generate_answer(context=context, question=question)
return answer
```
A program has two key methods, which you can edit to fit your needs.
**Your `__init__` method** declares the modules you will use. Here, `RAG` will use the built-in `Retrieve` for retrieval and `ChainOfThought` for generating answers. **DSPy** offers general-purpose modules that take the shape of _your own_ sub-tasks — and not pre-built functions for specific applications.
Modules that use the LM, like `ChainOfThought`, require a _signature_. That is a declarative spec that tells the module what it's expected to do. In this example, we use the short-hand signature notation `context, question -> answer` to tell `ChainOfThought` it will be given some `context` and a `question` and must produce an `answer`. We will discuss more advanced **[signatures](#3a-declaring-the-inputoutput-behavior-of-lms-with-dspysignature)** below.
**Your `forward` method** expresses any computation you want to do with your modules. In this case, we use the modules `self.retrieve` and `self.generate_answer` to search for some `context` and then use the `context` and `question` to generate the `answer`!
You can now either use this `RAG` program in **zero-shot mode**. Or **compile** it to obtain higher quality. Zero-shot usage is simple. Just define an instance of your program and then call it:
```python
rag = RAG() # zero-shot, uncompiled version of RAG
rag("what is the capital of France?").answer # -> "Paris"
```
The next section will discuss how to compile our simple `RAG` program. When we compile it, the **DSPy compiler** will annotate _demonstrations_ of its steps: (1) retrieval, (2) using context, and (3) using _chain-of-thought_ to answer questions. From these demonstrations, the **DSPy compiler** will make sure it produces an effective few-shot prompt that works well with your LM, retrieval model, and data. If you're working with small models, it'll finetune your model (instead of prompting) to do this task.
If you later decide you need another step in your pipeline, just add another module and compile again. Maybe add a module that takes the chat history into account during search?
## 3) Two Powerful Concepts: Signatures & Teleprompters
To make it possible to compile any program you write, **DSPy** introduces two simple concepts: Signatures and Teleprompters.
#### 3.a) Declaring the input/output behavior of LMs with `dspy.Signature`
When we assign tasks to LMs in **DSPy**, we specify the behavior we need as a **Signature**. A signature is a declarative specification of input/output behavior of a **DSPy module**.
Instead of investing effort into _how_ to get your LM to do a sub-task, signatures enable you to inform **DSPy** _what_ the sub-task is. Later, the **DSPy compiler** will figure out how to build a complex prompt for your large LM (or finetune your small LM) specifically for your signature, on your data, and within your pipeline.
A signature consists of three simple elements:
- A minimal description of the sub-task the LM is supposed to solve.
- A description of one or more input fields (e.g., input question) that will we will give to the LM.
- A description of one or more output fields (e.g., the question's answer) that we will expect from the LM.
We support two notations for expressing signatures. The **short-hand signature notation** is for quick development. You just provide your module (e.g., `dspy.ChainOfThought`) with a string with `input_field_name_1, ... -> output_field_name_1, ...` with the fields separated by commas.
In the `RAG` class earlier, we saw:
```python
self.generate_answer = dspy.ChainOfThought("context, question -> answer")
```
In many cases, this barebones signature is sufficient. However, sometimes you need more control. In these cases, we can use the full notation to express a more fully-fledged signature below.
```python
class GenerateSearchQuery(dspy.Signature):
"""Write a simple search query that will help answer a complex question."""
context = dspy.InputField(desc="may contain relevant facts")
question = dspy.InputField()
query = dspy.OutputField()
### inside your program's __init__ function
self.generate_answer = dspy.ChainOfThought(GenerateSearchQuery)
```
You can optionally provide a `prefix` and/or `desc` key for each input or output field to refine or constraint the behavior of modules using your signature. The description of the sub-task itself is specified as the docstring (i.e., `"""Write a simple..."""`).
#### 3.b) Asking **DSPy** to automatically optimize your program with `dspy.teleprompt.*`
After defining the `RAG` program, we can **compile** it. Compiling a program will update the parameters stored in each module. For large LMs, this is primarily in the form of creating and validating good demonstrations for inclusion in your prompt(s).
Compiling depends on three things: a (potentially tiny) training set, a metric for validation, and your choice of teleprompter from **DSPy**. **Teleprompters** are powerful optimizers (included in **DSPy**) that can learn to bootstrap and select effective prompts for the modules of any program. (The "tele-" in the name means "at a distance", i.e., automatic prompting at a distance.)
**DSPy** typically requires very minimal labeling. For example, our `RAG` pipeline may work well with just a handful of examples that contain a **question** and its (human-annotated) **answer**. Your pipeline may involve multiple complex steps: our basic `RAG` example includes a retrieved context, a chain of thought, and the answer. However, you only need labels for the initial question and the final answer. **DSPy** will bootstrap any intermediate labels needed to support your pipeline. If you change your pipeline in any way, the data bootstrapped will change accordingly!
```python
my_rag_trainset = [
dspy.Example(
question="Which award did Gary Zukav's first book receive?",
answer="National Book Award"
),
...
]
```
Second, define your validation logic, which will express some constraints on the behavior of your program or individual modules. For `RAG`, we might express a simple check like this:
```python
def validate_context_and_answer(example, pred, trace=None):
# check the gold label and the predicted answer are the same
answer_match = example.answer.lower() == pred.answer.lower()
# check the predicted answer comes from one of the retrieved contexts
context_match = any((pred.answer.lower() in c) for c in pred.context)
return answer_match and context_match
```
Different teleprompters offer various tradeoffs in terms of how much they optimize cost versus quality, etc. For `RAG`, we might use the simple teleprompter called `BootstrapFewShot`. To do so, we instantiate the teleprompter itself with a validation function `my_rag_validation_logic` and then compile against some training set `my_rag_trainset`.
```python
from dspy.teleprompt import BootstrapFewShot
teleprompter = BootstrapFewShot(metric=my_rag_validation_logic)
compiled_rag = teleprompter.compile(RAG(), trainset=my_rag_trainset)
```
If we now use `compiled_rag`, it will invoke our LM with rich prompts with few-shot demonstrations of chain-of-thought retrieval-augmented question answering on our data.
## 4) Documentation & Tutorials
While we work on new tutorials and documentation, please check out **[our intro notebook](intro.ipynb)**.
Or open it directly in free Google Colab: [<img align="center" src="https://colab.research.google.com/assets/colab-badge.svg" />](https://colab.research.google.com/github/stanfordnlp/dspy/blob/main/intro.ipynb)
###
<details>
<summary><h3 style="display: inline">Module Reference</h3></summary>
We have work-in-progress module documentation at [this PR](https://github.com/stanfordnlp/dspy/pull/93). Please let us know if anything there is unclear.
#### Language Model Clients
- [`dspy.OpenAI`](docs/language_models/language_models_client.md#openai)
- [`dspy.Cohere`](docs/language_models/language_models_client.md#cohere)
- [`dspy.TGI`](docs/language_models/language_models_client.md#tgi)
- [`dspy.VLLM`](docs/language_models/language_models_client.md#vllm)
#### Retrieval Model Clients
- [`dspy.ColBERTv2`](docs/retrieval/retrieval_models_client.md#colbertv2)
- [`dspy.AzureCognitiveSearch`](docs/retrieval/retrieval_models_client.md#azurecognitivesearch)
#### Signatures
- `dspy.Signature`
- `dspy.InputField`
- `dspy.OutputField`
#### Modules
- [`dspy.Predict`](docs/modules/modules.md#dspypredict)
- [`dspy.Retrieve`](docs/modules/modules.md#dspyretrieve)
- [`dspy.ChainOfThought`](docs/modules/modules.md#dspychainofthought)
- `dspy.majority` (functional self-consistency)
- `dspy.ProgramOfThought` [[see open PR]](https://github.com/stanfordnlp/dspy/pull/116)
- [`dspy.ReAct`](docs/modules/modules.md#dspyreact)
- [`dspy.MultiChainComparison`](docs/modules/modules.md#dspymultichaincomparison)
- `dspy.SelfCritique` [coming soon]
- `dspy.SelfRevision` [coming soon]
#### Teleprompters
- [`dspy.teleprompt.LabeledFewShot`](docs/Teleprompters/teleprompters.md#telepromptlabeledfewshot)
- [`dspy.teleprompt.BootstrapFewShot`](docs/Teleprompters/teleprompters.md#telepromptbootstrapfewshot)
- [`dspy.teleprompt.BootstrapFewShotWithRandomSearch`](docs/Teleprompters/teleprompters.md#telepromptbootstrapfewshotwithrandomsearch)
- `dspy.teleprompt.LabeledFinetune` [coming soon]
- [`dspy.teleprompt.BootstrapFinetune`](docs/Teleprompters/teleprompters.md#telepromptbootstrapfinetune)
- [`dspy.teleprompt.Ensemble`](docs/Teleprompters/teleprompters.md#telepromptensemble)
- `dspy.teleprompt.kNN` [coming soon]
</details>
###
<details>
<summary><h3 style="display: inline">Intro Tutorial [coming soon]</h3></summary>
**[Intro-01] Getting Started: High Quality Pipelined Prompts with Minimal Effort**
**[Intro-02] Using DSPy For Your Own Task: Building Blocks**
**[Intro-03] Adding Complexity: Multi-stage Programs**
**[Intro-04] Adding Complexity for Your Own Task: Design Patterns**
</details>
###
<details>
<summary><h3 style="display: inline">Advanced Demos [coming soon]</h3></summary>
**[Advanced-01] Long-Form QA & Programmatic Evaluation.**
**[Advanced-02] Programmatic Evaluation II & Dataset Creation.**
**[Advanced-03] Compiling & Teleprompters.**
**[Advanced-04] Extending DSPy with Modules or Teleprompters.**
**[Advanced-05]: Agents and General Tool Use in DSPy.**
**[Advanced-06]: Reproducibility, Saving Programs, and Advanced Caching.**
</details>
## 5) FAQ: Is DSPy right for me?
The **DSPy** philosophy and abstraction differ significantly from other libraries and frameworks, so it's usually straightforward to decide when **DSPy** is (or isn't) the right framework for your usecase.
If you're a NLP/AI researcher (or a practitioner exploring new pipelines or new tasks), the answer is generally an invariable **yes**. If you're a practitioner doing other things, please read on.
####
<details>
<summary><h4 style="display: inline">[5.a] DSPy vs. thin wrappers for prompts (OpenAI API, MiniChain, basic templating)</h4></summary>
In other words: _Why can't I just write my prompts directly as string templates?_ Well, for extremely simple settings, this _might_ work just fine. (If you're familiar with neural networks, this is like expressing a tiny two-layer NN as a Python for-loop. It kinda works.)
However, when you need higher quality (or manageable cost), then you need to iteratively explore multi-stage decomposition, improved prompting, data bootstrapping, careful finetuning, retrieval augmentation, and/or using smaller (or cheaper, or local) models. The true expressive power of building with foundation models lies in the interactions between these pieces. But every time you change one piece, you likely break (or weaken) multiple other components.
**DSPy** cleanly abstracts away (_and_ powerfully optimizes) the parts of these interactions that are external to your actual system design. It lets you focus on designing the module-level interactions: the _same program_ expressed in 10 or 20 lines of **DSPy** can easily be compiled into multi-stage instructions for `GPT-4`, detailed prompts for `Llama2-13b`, or finetunes for `T5-base`.
Oh, and you wouldn't need to maintain long, brittle, model-specific strings at the core of your project anymore.
</details>
####
<details>
<summary><h4 style="display: inline">[5.b] DSPy vs. application development libraries like LangChain, LlamaIndex</h4></summary>
> _Note: If you use LangChain as a thin wrapper around your own prompt strings, refer to answer [5.a] instead._
LangChain and LlamaIndex are popular libraries that target high-level application development with LMs. They offer many _batteries-included_, pre-built application modules that plug in with your data or configuration. In practice, indeed, many usecases genuinely _don't need_ any special components. If you'd be happy to use someone's generic, off-the-shelf prompt for question answering over PDFs or standard text-to-SQL as long as it's easy to set up on your data, then you will probably find a very rich ecosystem in these libraries.
Unlike these libraries, **DSPy** doesn't internally contain hand-crafted prompts that target specific applications you can build. Instead, **DSPy** introduces a very small set of much more powerful and general-purpose modules _that can learn to prompt (or finetune) your LM within your pipeline on your data_.
**DSPy** offers a whole different degree of modularity: when you change your data, make tweaks to your program's control flow, or change your target LM, the **DSPy compiler** can map your program into a new set of prompts (or finetunes) that are optimized specifically for this pipeline. Because of this, you may find that **DSPy** obtains the highest quality for your task, with the least effort, provided you're willing to implement (or extend) your own short program. In short, **DSPy** is for when you need a lightweight but automatically-optimizing programming model — not a library of predefined prompts and integrations.
If you're familiar with neural networks:
> This is like the difference between PyTorch (i.e., representing **DSPy**) and HuggingFace Transformers (i.e., representing the higher-level libraries). If you simply want to use off-the-shelf `BERT-base-uncased` or `GPT2-large` or apply minimal finetuning to them, HF Transformers makes it very straightforward. If, however, you're looking to build your own architecture (or extend an existing one significantly), you have to quickly drop down into something much more modular like PyTorch. Luckily, HF Transformers _is_ implemented in backends like PyTorch. We are similarly excited about high-level wrapper around **DSPy** for common applications. If this is implemented using **DSPy**, your high-level application can also adapt significantly to your data in a way that static prompt chains won't. Please [open an issue](https://github.com/stanfordnlp/dspy/issues/new) if this is something you want to help with.
</details>
####
<details>
<summary><h4 style="display: inline">[5.c] DSPy vs. generation control libraries like Guidance, LMQL, RELM, Outlines</h4></summary>
Guidance, LMQL, RELM, and Outlines are all exciting new libraries for controlling the individual completions of LMs, e.g., if you want to enforce JSON output schema or constrain sampling to a particular regular expression.
This is very useful in many settings, but it's generally focused on low-level, structured control of a single LM call. It doesn't help ensure the JSON (or structured output) you get is going to be correct or useful for your task.
In contrast, **DSPy** automatically optimizes the prompts in your programs to align them with various task needs, which may also include producing valid structured ouputs. That said, we are considering allowing **Signatures** in **DSPy** to express regex-like constraints that are implemented by these libraries.
</details>
## Contributors & Acknowledgements
**DSPy** is led by **Omar Khattab** at Stanford NLP with **Chris Potts** and **Matei Zaharia**.
Key contributors and team members include **Arnav Singhvi**, **Paridhi Maheshwari**, **Keshav Santhanam**, **Sri Vardhamanan**, **Eric Zhang**, **Hanna Moazam**, **Thomas Joshi**, **Saiful Haq**, and **Ashutosh Sharma**.
**DSPy** includes important contributions from **Rick Battle** and **Igor Kotenkov**. It reflects discussions with **Lisa Li**, **David Hall**, **Ashwin Paranjape**, **Heather Miller**, **Chris Manning**, **Percy Liang**, and many others.
The **DSPy** logo is designed by **Chuyi Zhang**.
## 📜 Citation & Reading More
To stay up to date or learn more, follow [@lateinteraction](https://twitter.com/lateinteraction) on Twitter.
If you use DSPy or DSP in a research paper, please cite our work as follows:
```
@article{khattab2023dspy,
title={DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines},
author={Khattab, Omar and Singhvi, Arnav and Maheshwari, Paridhi and Zhang, Zhiyuan and Santhanam, Keshav and Vardhamanan, Sri and Haq, Saiful and Sharma, Ashutosh and Joshi, Thomas T. and Moazam, Hanna and Miller, Heather and Zaharia, Matei and Potts, Christopher},
journal={arXiv preprint arXiv:2310.03714},
year={2023}
}
@article{khattab2022demonstrate,
title={Demonstrate-Search-Predict: Composing Retrieval and Language Models for Knowledge-Intensive {NLP}},
author={Khattab, Omar and Santhanam, Keshav and Li, Xiang Lisa and Hall, David and Liang, Percy and Potts, Christopher and Zaharia, Matei},
journal={arXiv preprint arXiv:2212.14024},
year={2022}
}
```
You can also read more about the evolution of the framework from Demonstrate-Search-Predict to DSPy:
* [**DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines**](https://arxiv.org/abs/2310.03714) (Academic Paper, Oct 2023)

Step 2: ⌨️ Coding

Create docs/datasets/dataset.md with contents:
• Create a new markdown file named `dataset.md` in the `docs/datasets` directory.
• Start by providing a brief overview of the `dataset.py` file, explaining its purpose and functionality.
• Describe the `Dataset` class, its methods, and their usage. Explain the purpose of each method and how it contributes to the overall functionality of the class.
• Draw from the README.md file to explain the philosophical usage/purpose of the `Dataset` class.
• Include examples of how to use the `Dataset` class. These examples can be drawn from the jupyter notebooks.
Create docs/datasets/gsm8k.md with contents:
• Create a new markdown file named `gsm8k.md` in the `docs/datasets` directory.
• Start by providing a brief overview of the `gsm8k.py` file, explaining its purpose and functionality.
• Describe the `GSM8K` class, its methods, and their usage. Explain the purpose of each method and how it contributes to the overall functionality of the class.
• Draw from the README.md file to explain the philosophical usage/purpose of the `GSM8K` class.
• Include examples of how to use the `GSM8K` class. These examples can be drawn from the jupyter notebooks.
Create docs/datasets/hotpotqa.md with contents:
• Create a new markdown file named `hotpotqa.md` in the `docs/datasets` directory.
• Start by providing a brief overview of the `hotpotqa.py` file, explaining its purpose and functionality.
• Describe the `HotPotQA` class, its methods, and their usage. Explain the purpose of each method and how it contributes to the overall functionality of the class.
• Draw from the README.md file to explain the philosophical usage/purpose of the `HotPotQA` class.
• Include examples of how to use the `HotPotQA` class. These examples can be drawn from the jupyter notebooks.
Create docs/datasets/colors.md with contents:
• Create a new markdown file named `colors.md` in the `docs/datasets` directory.
• Start by providing a brief overview of the `colors.py` file, explaining its purpose and functionality.
• Describe the `Colors` class, its methods, and their usage. Explain the purpose of each method and how it contributes to the overall functionality of the class.
• Draw from the README.md file to explain the philosophical usage/purpose of the `Colors` class.
• Include examples of how to use the `Colors` class. These examples can be drawn from the jupyter notebooks.

Step 3: 🔁 Code Review

I have finished reviewing the code for completeness. I did not find errors for sweep/ensure_datasets_in_the_dspy_folder_has_d.

🎉 Latest improvements to Sweep:

  • We just released a dashboard to track Sweep's progress on your issue in real-time, showing every stage of the process – from search to planning and coding.
  • Sweep uses OpenAI's latest Assistant API to plan code changes and modify code! This is 3x faster and significantly more reliable as it allows Sweep to edit code and validate the changes in tight iterations, the same way as a human would.
  • Try using the GitHub issues extension to create Sweep issues directly from your editor! GitHub Issues and Pull Requests.

💡 To recreate the pull request edit the issue title or description. To tweak the pull request, leave a comment on the pull request.
Join Our Discord

from dspy.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.