bigstorm's picture
Upload folder using huggingface_hub
c7095eb verified
metadata
license: llama3
tags:
  - function-calling

BigStorm - ExLLamaV2 (Exl2) Quantization

  • 8.0 bpw target
  • 8 head bits

Enjoy! Raise an issue if you'd like other BPW levels. Base Model Card Follows:

FireFunction V2: Fireworks Function Calling Model

Try on Fireworks | API Docs | Demo App | Discord

firefunction

FireFunction is a state-of-the-art function calling model with a commercially viable license. View detailed info in our announcement blog. Key info and highlights:

Comparison with other models:

  • Competitive with GPT-4o at function-calling, scoring 0.81 vs 0.80 on a medley of public evaluations
  • Trained on Llama 3 and retains Llama 3’s conversation and instruction-following capabilities, scoring 0.84 vs Llama 3’s 0.89 on MT bench
  • Significant quality improvements over FireFunction v1 across the broad range of metrics

General info:

🐾 Successor of the FireFunction model

🔆 Support of parallel function calling (unlike FireFunction v1) and good instruction following

💡 Hosted on the Fireworks platform at < 10% of the cost of GPT 4o and 2x the speed

Intended Use and Limitations

Supported usecases

The model was tuned to perfom well on a range of usecases including:

  • general instruction following
  • multi-turn chat mixing vanilla messages with function calls
  • single- and parallel function calling
  • up to 20 function specs supported at once
  • structured information extraction

The model has an 8k context window, like Llama 3

Out-of-Scope Use

The model was not optimized for the following use cases:

  • 100+ function specs
  • nested function calling

Metrics

Benchmark Firefunction v1 Firefunction v2 Llama 3 70b Instruct Gpt-4o
Gorilla simple 0.91 0.94 0.925 0.88
Gorilla multiple_function 0.92 0.91 0.86 0.91
Gorilla parallel_function 0 0.9 0.86 0.89
Gorilla parallel_multiple_function 0 0.8 0.615 0.72
Nexus parallel 0.38 0.53 0.3 0.47
Mtbench 0.73 0.84 0.89 0.93
Average 0.49 0.82 0.74 0.8

Example Usage

See documentation for more detail.

from transformers import AutoModelForCausalLM, AutoTokenizer
import json
from datetime import datetime

device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained("fireworks-ai/firefunction-v2", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("fireworks-ai/firefunction-v2")

function_spec = [
    {
        "name": "get_stock_price",
        "description": "Get the current stock price",
        "parameters": {
            "type": "object",
            "properties": {
                "symbol": {
                    "type": "string",
                    "description": "The stock symbol, e.g. AAPL, GOOG"
                }
            },
            "required": [
                "symbol"
            ]
        }
    },
    {
        "name": "check_word_anagram",
        "description": "Check if two words are anagrams of each other",
        "parameters": {
            "type": "object",
            "properties": {
                "word1": {
                    "type": "string",
                    "description": "The first word"
                },
                "word2": {
                    "type": "string",
                    "description": "The second word"
                }
            },
            "required": [
                "word1",
                "word2"
            ]
        }
    }
]
functions = json.dumps(function_spec, indent=4)

messages = [
    {'role': 'system', 'content': 'You are a helpful assistant with access to functions. Use them if required.'},
    {'role': 'user', 'content': 'Hi, can you tell me the current stock price of google and netflix?'}
]

now = datetime.now().strftime('%Y-%m-%d %H:%M:%S')

model_inputs = tokenizer.apply_chat_template(messages, functions=functions, datetime=now, return_tensors="pt").to(model.device)

generated_ids = model.generate(model_inputs, max_new_tokens=128)
decoded = tokenizer.batch_decode(generated_ids)
print(decoded[0])

Resources