firefunction-v2-6.0bpw-8hb-exl2

BigStorm - ExLLamaV2 (Exl2) Quantization

• 6.0 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

<img src="https://cdn-uploads.huggingface.co/production/uploads/64b6f3a72f5a966b9722de88/nJNtxLzWswBDKK1iOZblb.png" alt="firefunction" width="400"/>

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

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

• Fireworks discord with function calling channel
• Documentation
• Demo app
• Try in Fireworks prompt playground UI