LogoLogo
Continuum WebsiteContinuum ApplicationsContinuum KnowledgeAxolotl Platform
Continuum Knowledge
Continuum Knowledge
  • Continuum
  • Data
    • Datasets
      • Pre Training Data
      • Types of Fine Tuning
      • Self Instruct Paper
      • Self-Alignment with Instruction Backtranslation
      • Systematic Evaluation of Instruction-Tuned Large Language Models on Open Datasets
      • Instruction Tuning
      • Instruction Fine Tuning - Alpagasus
      • Less is More For Alignment
      • Enhanced Supervised Fine Tuning
      • Visualising Data using t-SNE
      • UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
      • Training and Evaluation Datasets
      • What is perplexity?
  • MODELS
    • Foundation Models
      • The leaderboard
      • Foundation Models
      • LLama 2 - Analysis
      • Analysis of Llama 3
      • Llama 3.1 series
      • Google Gemini 1.5
      • Platypus: Quick, Cheap, and Powerful Refinement of LLMs
      • Mixtral of Experts
      • Mixture-of-Agents (MoA)
      • Phi 1.5
        • Refining the Art of AI Training: A Deep Dive into Phi 1.5's Innovative Approach
      • Phi 2.0
      • Phi-3 Technical Report
  • Training
    • The Fine Tuning Process
      • Why fine tune?
        • Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?
        • Explanations in Fine Tuning
      • Tokenization
        • Tokenization Is More Than Compression
        • Tokenization - SentencePiece
        • Tokenization explore
        • Tokenizer Choice For LLM Training: Negligible or Crucial?
        • Getting the most out of your tokenizer for pre-training and domain adaptation
        • TokenMonster
      • Parameter Efficient Fine Tuning
        • P-Tuning
          • The Power of Scale for Parameter-Efficient Prompt Tuning
        • Prefix-Tuning: Optimizing Continuous Prompts for Generation
        • Harnessing the Power of PEFT: A Smarter Approach to Fine-tuning Pre-trained Models
        • What is Low-Rank Adaptation (LoRA) - explained by the inventor
        • Low Rank Adaptation (Lora)
        • Practical Tips for Fine-tuning LMs Using LoRA (Low-Rank Adaptation)
        • QLORA: Efficient Finetuning of Quantized LLMs
        • Bits and Bytes
        • The Magic behind Qlora
        • Practical Guide to LoRA: Tips and Tricks for Effective Model Adaptation
        • The quantization constant
        • QLORA: Efficient Finetuning of Quantized Language Models
        • QLORA and Fine-Tuning of Quantized Language Models (LMs)
        • ReLoRA: High-Rank Training Through Low-Rank Updates
        • SLoRA: Federated Parameter Efficient Fine-Tuning of Language Models
        • GaLora: Memory-Efficient LLM Training by Gradient Low-Rank Projection
      • Hyperparameters
        • Batch Size
        • Padding Tokens
        • Mixed precision training
        • FP8 Formats for Deep Learning
        • Floating Point Numbers
        • Batch Size and Model loss
        • Batch Normalisation
        • Rethinking Learning Rate Tuning in the Era of Language Models
        • Sample Packing
        • Gradient accumulation
        • A process for choosing the learning rate
        • Learning Rate Scheduler
        • Checkpoints
        • A Survey on Efficient Training of Transformers
        • Sequence Length Warmup
        • Understanding Training vs. Evaluation Data Splits
        • Cross-entropy loss
        • Weight Decay
        • Optimiser
        • Caching
      • Training Processes
        • Extending the context window
        • PyTorch Fully Sharded Data Parallel (FSDP)
        • Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
        • YaRN: Efficient Context Window Extension of Large Language Models
        • Sliding Window Attention
        • LongRoPE
        • Reinforcement Learning
        • An introduction to reinforcement learning
        • Reinforcement Learning from Human Feedback (RLHF)
        • Direct Preference Optimization: Your Language Model is Secretly a Reward Model
  • INFERENCE
    • Why is inference important?
      • Grouped Query Attention
      • Key Value Cache
      • Flash Attention
      • Flash Attention 2
      • StreamingLLM
      • Paged Attention and vLLM
      • TensorRT-LLM
      • Torchscript
      • NVIDIA L40S GPU
      • Triton Inference Server - Introduction
      • Triton Inference Server
      • FiDO: Fusion-in-Decoder optimised for stronger performance and faster inference
      • Is PUE a useful measure of data centre performance?
      • SLORA
  • KNOWLEDGE
    • Vector Databases
      • A Comprehensive Survey on Vector Databases
      • Vector database management systems: Fundamental concepts, use-cases, and current challenges
      • Using the Output Embedding to Improve Language Models
      • Decoding Sentence-BERT
      • ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
      • SimCSE: Simple Contrastive Learning of Sentence Embeddings
      • Questions Are All You Need to Train a Dense Passage Retriever
      • Improving Text Embeddings with Large Language Models
      • Massive Text Embedding Benchmark
      • RocketQAv2: A Joint Training Method for Dense Passage Retrieval and Passage Re-ranking
      • LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders
      • Embedding and Fine-Tuning in Neural Language Models
      • Embedding Model Construction
      • Demystifying Embedding Spaces using Large Language Models
      • Fine-Tuning Llama for Multi-Stage Text Retrieval
      • Large Language Model Based Text Augmentation Enhanced Personality Detection Model
      • One Embedder, Any Task: Instruction-Finetuned Text Embeddings
      • Vector Databases are not the only solution
      • Knowledge Graphs
        • Harnessing Knowledge Graphs to Elevate AI: A Technical Exploration
        • Unifying Large Language Models and Knowledge Graphs: A Roadmap
      • Approximate Nearest Neighbor (ANN)
      • High Dimensional Data
      • Principal Component Analysis (PCA)
      • Vector Similarity Search - HNSW
      • FAISS (Facebook AI Similarity Search)
      • Unsupervised Dense Retrievers
    • Retrieval Augmented Generation
      • Retrieval-Augmented Generation for Large Language Models: A Survey
      • Fine-Tuning or Retrieval?
      • Revolutionising Information Retrieval: The Power of RAG in Language Models
      • A Survey on Retrieval-Augmented Text Generation
      • REALM: Retrieval-Augmented Language Model Pre-Training
      • Retrieve Anything To Augment Large Language Models
      • Generate Rather Than Retrieve: Large Language Models Are Strong Context Generators
      • Active Retrieval Augmented Generation
      • DSPy: LM Assertions: Enhancing Language Model Pipelines with Computational Constraints
      • DSPy: Compiling Declarative Language Model Calls
      • DSPy: In-Context Learning for Extreme Multi-Label Classification
      • Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs
      • HYDE: Revolutionising Search with Hypothetical Document Embeddings
      • Enhancing Recommender Systems with Large Language Model Reasoning Graphs
      • Retrieval Augmented Generation (RAG) versus fine tuning
      • RAFT: Adapting Language Model to Domain Specific RAG
      • Summarisation Methods and RAG
      • Lessons Learned on LLM RAG Solutions
      • Stanford: Retrieval Augmented Language Models
      • Overview of RAG Approaches with Vector Databases
      • Mastering Chunking in Retrieval-Augmented Generation (RAG) Systems
    • Semantic Routing
    • Resource Description Framework (RDF)
  • AGENTS
    • What is agency?
      • Rephrase and Respond: Let Large Language Models Ask Better Questions for Themselves
      • Types of Agents
      • The risk of AI agency
      • Understanding Personality in Large Language Models: A New Frontier in AI Psychology
      • AI Agents - Reasoning, Planning, and Tool Calling
      • Personality and Brand
      • Agent Interaction via APIs
      • Bridging Minds and Machines: The Legacy of Newell, Shaw, and Simon
      • A Survey on Language Model based Autonomous Agents
      • Large Language Models as Agents
      • AI Reasoning: A Deep Dive into Chain-of-Thought Prompting
      • Enhancing AI Reasoning with Self-Taught Reasoner (STaR)
      • Exploring the Frontier of AI: The "Tree of Thoughts" Framework
      • Toolformer: Revolutionising Language Models with API Integration - An Analysis
      • TaskMatrix.AI: Bridging Foundational AI Models with Specialised Systems for Enhanced Task Completion
      • Unleashing the Power of LLMs in API Integration: The Rise of Gorilla
      • Andrew Ng's presentation on AI agents
      • Making AI accessible with Andrej Karpathy and Stephanie Zhan
  • Regulation and Ethics
    • Regulation and Ethics
      • Privacy
      • Detecting AI Generated content
      • Navigating the IP Maze in AI: The Convergence of Blockchain, Web 3.0, and LLMs
      • Adverse Reactions to generative AI
      • Navigating the Ethical Minefield: The Challenge of Security in Large Language Models
      • Navigating the Uncharted Waters: The Risks of Autonomous AI in Military Decision-Making
  • DISRUPTION
    • Data Architecture
      • What is a data pipeline?
      • What is Reverse ETL?
      • Unstructured Data and Generatve AI
      • Resource Description Framework (RDF)
      • Integrating generative AI with the Semantic Web
    • Search
      • BM25 - Search Engine Ranking Function
      • BERT as a reranking engine
      • BERT and Google
      • Generative Engine Optimisation (GEO)
      • Billion-scale similarity search with GPUs
      • FOLLOWIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions
      • Neural Collaborative Filtering
      • Federated Neural Collaborative Filtering
      • Latent Space versus Embedding Space
      • Improving Text Embeddings with Large Language Models
    • Recommendation Engines
      • On Interpretation and Measurement of Soft Attributes for Recommendation
      • A Survey on Large Language Models for Recommendation
      • Model driven recommendation systems
      • Recommender AI Agent: Integrating Large Language Models for Interactive Recommendations
      • Foundation Models for Recommender Systems
      • Exploring the Impact of Large Language Models on Recommender Systems: An Extensive Review
      • AI driven recommendations - harming autonomy?
    • Logging
      • A Taxonomy of Anomalies in Log Data
      • Deeplog
      • LogBERT: Log Anomaly Detection via BERT
      • Experience Report: Deep Learning-based System Log Analysis for Anomaly Detection
      • Log-based Anomaly Detection with Deep Learning: How Far Are We?
      • Deep Learning for Anomaly Detection in Log Data: A Survey
      • LogGPT
      • Adaptive Semantic Gate Networks (ASGNet) for log-based anomaly diagnosis
  • Infrastructure
    • The modern data centre
      • Enhancing Data Centre Efficiency: Strategies to Improve PUE
      • TCO of NVIDIA GPUs and falling barriers to entry
      • Maximising GPU Utilisation with Kubernetes and NVIDIA GPU Operator
      • Data Centres
      • Liquid Cooling
    • Servers and Chips
      • The NVIDIA H100 GPU
      • NVIDIA H100 NVL
      • Lambda Hyperplane 8-H100
      • NVIDIA DGX Servers
      • NVIDIA DGX-2
      • NVIDIA DGX H-100 System
      • NVLink Switch
      • Tensor Cores
      • NVIDIA Grace Hopper Superchip
      • NVIDIA Grace CPU Superchip
      • NVIDIA GB200 NVL72
      • Hopper versus Blackwell
      • HGX: High-Performance GPU Platforms
      • ARM Chips
      • ARM versus x86
      • RISC versus CISC
      • Introduction to RISC-V
    • Networking and Connectivity
      • Infiniband versus Ethernet
      • NVIDIA Quantum InfiniBand
      • PCIe (Peripheral Component Interconnect Express)
      • NVIDIA ConnectX InfiniBand adapters
      • NVMe (Non-Volatile Memory Express)
      • NVMe over Fabrics (NVMe-oF)
      • NVIDIA Spectrum-X
      • NVIDIA GPUDirect
      • Evaluating Modern GPU Interconnect
      • Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)
      • Next-generation networking in AI environments
      • NVIDIA Collective Communications Library (NCCL)
    • Data and Memory
      • NVIDIA BlueField Data Processing Units (DPUs)
      • Remote Direct Memory Access (RDMA)
      • High Bandwidth Memory (HBM3)
      • Flash Memory
      • Model Requirements
      • Calculating GPU memory for serving LLMs
      • Transformer training costs
      • GPU Performance Optimisation
    • Libraries and Complements
      • NVIDIA Base Command
      • NVIDIA AI Enterprise
      • CUDA - NVIDIA GTC 2024 presentation
      • RAPIDs
      • RAFT
    • Vast Data Platform
      • Vast Datastore
      • Vast Database
      • Vast Data Engine
      • DASE (Disaggregated and Shared Everything)
      • Dremio and VAST Data
    • Storage
      • WEKA: A High-Performance Storage Solution for AI Workloads
      • Introduction to NVIDIA GPUDirect Storage (GDS)
        • GDS cuFile API
      • NVIDIA Magnum IO GPUDirect Storage (GDS)
      • Vectors in Memory
Powered by GitBook
LogoLogo

Continuum - Accelerated Artificial Intelligence

  • Continuum Website
  • Axolotl Platform

Copyright Continuum Labs - 2023

On this page
  • The Challenge with LLMs and API Calls
  • Gorilla: A New Hope
  • Advancements and Insights
  • Conclusion: A Step Towards Responsible AI Development
  • Practical Application
  • Calling an API to make a airflight booking
  • Conclusion

Was this helpful?

  1. AGENTS
  2. What is agency?

Unleashing the Power of LLMs in API Integration: The Rise of Gorilla

PreviousTaskMatrix.AI: Bridging Foundational AI Models with Specialised Systems for Enhanced Task CompletionNextAndrew Ng's presentation on AI agents

Last updated 11 months ago

Was this helpful?

This May 2023 paper reveals "Gorilla," a fine-tuned model designed to bridge the divide between LLMs and their ability to make accurate API calls.

The Challenge with LLMs and API Calls

The study spotlights the limitations of current LLMs in generating precise input arguments for APIs and their propensity to "hallucinate" incorrect API usage.

This issue not only hampers the efficiency of LLMs but also restricts their practical application in real-world scenarios, where the ability to interact seamlessly with various tools and platforms through APIs is indispensable.

Gorilla: A New Hope

To surmount these challenges, the research introduces "Gorilla," a model that leverages the LLaMA architecture and surpasses GPT-4 in crafting API calls with minimised errors.

Gorilla's integration with a document retriever stands out as a hallmark of its design, allowing it to adapt to real-time updates or changes in API documentation, thereby significantly boosting its reliability and flexibility.

The Backbone: APIBench Dataset

A key component of this study is the APIBench dataset, an extensive collection of APIs from leading platforms like HuggingFace, TorchHub, and TensorHub.

This dataset plays a crucial role in evaluating the model’s capability to make functional API calls, setting a new benchmark for assessing LLMs in this domain.

Fine-tuning Gorilla: A Detailed Look

Gorilla undergoes fine-tuning on the APIBench dataset with a keen focus on enhancing its document retrieval capabilities.

This process not only sharpens Gorilla's accuracy in API functionality but also significantly reduces the instances of hallucinations, marking a considerable advancement in the precision of API usage by LLMs.

Advancements and Insights

Gorilla’s introduction is a move forward in enabling LLMs to interact with APIs more effectively.

By integrating a document retriever and undergoing evaluation on a dataset, Gorilla showcases notable improvements in accuracy and reliability over its predecessors, such as GPT-4.

This is especially evident in the domain of program synthesis using API calls, where Gorilla's ability to understand and reason about constraints represents a significant improvement.

Conclusion: A Step Towards Responsible AI Development

The research encapsulates a significant breakthrough in the realm of LLMs and their interaction with APIs.

Gorilla not only enhances the capability of LLMs to make precise and reliable API calls but also addresses the ethical and social implications of deploying such technologies.

By releasing an extensive dataset to the community, the researchers underscore their commitment to fostering a more equitable and optimized use of machine learning, paving the way for future advancements in this field.

Practical Application

To demonstrate how the concepts discussed in the paper could be operationalized through code, let's create a simplified version of what might resemble the Gorilla model's approach to making API calls. This example will focus on the following key aspects:

  1. API Call Generation: Writing a function that simulates generating an API call based on input instructions.

  2. Integration with Document Retriever: Simulating how Gorilla integrates with a document retriever to adapt to API documentation changes.

  3. Handling Constraints: Demonstrating how the model might handle constraints when making API calls.

Since the detailed implementation of Gorilla, including its fine-tuning on LLaMA and integration with a sophisticated document retrieval system, is beyond the scope of this example, we'll use Python pseudocode for illustrative purposes.

API Call Generation

Let's start by simulating how Gorilla generates an API call. Assume we have a simplified function that takes an instruction and selects an API based on the task description.

def generate_api_call(instruction):
    # Example instruction: "Classify an image into a category."
    # This is a simplified version of how Gorilla might interpret the instruction
    # and decide which API to use based on its training.
    if "classify an image" in instruction.lower():
        return "POST", "https://api.example.com/v1/image/classification"
    elif "translate text" in instruction.lower():
        return "POST", "https://api.example.com/v1/text/translation"
    else:
        return "GET", "https://api.example.com/v1/unknown/task"

instruction = "Classify an image into a category."
method, url = generate_api_call(instruction)
print(f"API Call Generated: {method} {url}")

Integration with Document Retriever

Next, let's simulate how Gorilla uses a document retriever to ensure it's using the most current version of the API documentation.

def retrieve_latest_api_doc(api_url):
    # This function simulates the retrieval of the latest API documentation.
    # In practice, this could involve querying a database or using a search engine.
    # Here, we'll simply return a simulated documentation snippet.
    documentation = {
        "https://api.example.com/v1/image/classification": {"version": "v1", "method": "POST"},
        "https://api.example.com/v1/text/translation": {"version": "v2", "method": "POST"}  # Note the version update
    }
    return documentation.get(api_url, {})

# Assuming the instruction is for text translation, which has an updated API version
instruction = "Translate text from English to French."
method, url = generate_api_call(instruction)
doc = retrieve_latest_api_doc(url)

# Update the API call based on the retrieved documentation
if doc:  # If documentation was found
    method = doc["method"]
    url = url.replace("v1", doc["version"])  # Update URL with the correct version

print(f"Updated API Call: {method} {url}")

Handling Constraints

Finally, let's demonstrate how constraints might be handled. Assume we want to enforce a constraint that the API call must not exceed a certain response time.

def make_api_call_with_constraints(url, method, max_response_time=500):
    # This function simulates making an API call while respecting a maximum response time constraint.
    # For the sake of this example, we'll randomly determine if the constraint is met.
    import random
    response_time = random.randint(100, 600)  # Simulate response time in milliseconds
    
    if response_time <= max_response_time:
        return True, f"API call successful with response time {response_time}ms."
    else:
        return False, f"API call exceeded maximum response time with {response_time}ms."

# Making an API call with constraints
success, message = make_api_call_with_constraints(url, method)
print(message)

This code provides a basic framework to understand how Gorilla might operate in generating and executing API calls, integrating with document retrievers, and handling constraints. It's important to note that the actual implementation of Gorilla would involve complex models, extensive datasets, and sophisticated algorithms not covered in this simplified example.

Calling an API to make a airflight booking

In this demonstration, we'll simulate how an LLM, inspired by the Gorilla concept, would interact with a fictional airline's API to book a flight. Let's start by outlining a brief API documentation for "SkyHigh Airlines" and then proceed with how the LLM would generate and execute the API call.

SkyHigh Airlines API Documentation

Endpoint: POST /api/bookFlight

This endpoint is used to book a flight with SkyHigh Airlines.

Request Parameters:

  • origin (string): The departure city code (e.g., "NYC").

  • destination (string): The arrival city code (e.g., "LAX").

  • date (string): The departure date in YYYY-MM-DD format.

  • class (string): The class of service ("economy", "business", "first").

  • passengers (int): Number of passengers.

Response:

  • confirmationNumber (string): The booking confirmation number.

  • details (object): Object containing booking details including flight number, departure time, and total cost.

Model Interaction

For the LLM interaction, we'll simulate how the LLM, named "GorillaFlight", decides on the appropriate API call based on a user's instruction to book a flight.

Then, we'll show a simple Python function that represents how the LLM might internally generate and execute the API call.

Step 1: Understanding the Instruction

User instruction to the LLM: "Book an economy class flight from New York City to Los Angeles for two passengers on March 15th, 2024."

Step 2: Generating the API Call

Based on the instruction, "GorillaFlight" identifies the key parameters needed for the API call: origin, destination, date, class, and passengers.

def generate_api_call_for_flight(instruction):
    # Simplified NLP processing to extract information from the instruction
    # In practice, this would involve complex parsing and understanding
    origin = "NYC"  # Extracted from "New York City"
    destination = "LAX"  # Extracted from "Los Angeles"
    date = "2024-03-15"  # Extracted from "March 15th, 2024"
    flight_class = "economy"  # Extracted from "economy class"
    passengers = 2  # Extracted from "two passengers"
    
    api_url = "https://api.skyhighairlines.com/api/bookFlight"
    payload = {
        "origin": origin,
        "destination": destination,
        "date": date,
        "class": flight_class,
        "passengers": passengers
    }
    
    return api_url, payload

api_url, payload = generate_api_call_for_flight("Book an economy class flight from New York City to Los Angeles for two passengers on March 15th, 2024.")

Step 3: Making the API Call

In this step, we simulate the API call using the generated information.

def make_flight_booking(api_url, payload):
    # This function simulates making the actual API call.
    # For demonstration, we'll return a mock response instead of performing a real HTTP request.
    
    mock_response = {
        "confirmationNumber": "SKY123456",
        "details": {
            "flightNumber": "SKY789",
            "departureTime": "2024-03-15T08:00:00",
            "totalCost": 560.00
        }
    }
    
    print(f"Booking successful. Confirmation Number: {mock_response['confirmationNumber']}")
    print(f"Flight Details: {mock_response['details']}")

# Simulate making the API call
make_flight_booking(api_url, payload)

Conclusion

This simplified simulation showcases how an LLM like "GorillaFlight" could interpret a natural language instruction, generate a structured API call to book a flight with SkyHigh Airlines, and handle the response.

The actual implementation would involve advanced techniques for parsing and understanding the instructions, secure and efficient mechanisms for making HTTP requests, and robust error handling to manage various edge cases and API response scenarios. This is just a fun example of how the technology will eventually be used.

LogoGorilla: Large Language Model Connected with Massive APIsarXiv.org
Gorilla; Calling APIs
Page cover image