Instructions to use fireworks-ai/firefunction-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use fireworks-ai/firefunction-v1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="fireworks-ai/firefunction-v1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("fireworks-ai/firefunction-v1")
model = AutoModelForCausalLM.from_pretrained("fireworks-ai/firefunction-v1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use fireworks-ai/firefunction-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "fireworks-ai/firefunction-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "fireworks-ai/firefunction-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/fireworks-ai/firefunction-v1

SGLang

How to use fireworks-ai/firefunction-v1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "fireworks-ai/firefunction-v1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "fireworks-ai/firefunction-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "fireworks-ai/firefunction-v1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "fireworks-ai/firefunction-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use fireworks-ai/firefunction-v1 with Docker Model Runner:
```
docker model run hf.co/fireworks-ai/firefunction-v1
```

raymond513 commited on Feb 20, 2024

Commit

1a46f92

verified ·

1 Parent(s): 7e2f8d4

Update README.md

Browse files

Files changed (1) hide show

README.md +10 -10

README.md CHANGED Viewed

@@ -8,9 +8,17 @@ tags:
 <img src="https://cdn-uploads.huggingface.co/production/uploads/64b6f3a72f5a966b9722de88/12mfdeAJzW1NdKrN_J--L.png" alt="firefunction" width="400"/>
-FireFunction is a state-of-the-art function calling model with a commercially viable license.
-💡 The model is also hosted on the [Fireworks](https://fireworks.ai/models/fireworks/firefunction-v1) platform. It's free during a limited beta period and hosted to be about 4x the speed of GPT-4, generating ~110 tokens/sec! The model is also API compatible with [OpenAI function calling](https://platform.openai.com/docs/guides/function-calling).
 ```sh
 OPENAI_API_BASE=https://api.fireworks.ai/inference/v1
 OPENAI_API_KEY=<YOUR_FIREWORKS_API_KEY>
@@ -28,14 +36,6 @@ MODEL=accounts/fireworks/models/firefunction-v1
 # Intended Use and Limitations
-### Key Highlights
-⭐️ Near GPT-4 level quality for real-world use cases of structured information generation and routing decision-making
-💨 Blazing fast speed. Inference speeds are roughly 4x that of GPT-4 when using FireFunction hosted on the Fireworks platform
-🔄 Support for "any" paramter in tool_choice. Firefunction is the only model that we're aware that supports an option for the model to always choose a function - particularly helpful for routing use cases
 ### Primary Use
 Although the model was trained on a variety of tasks, it performs best on:
  * single-turn request routing to a function picked from a pool of up to 20 function specs.

 <img src="https://cdn-uploads.huggingface.co/production/uploads/64b6f3a72f5a966b9722de88/12mfdeAJzW1NdKrN_J--L.png" alt="firefunction" width="400"/>
+FireFunction is a state-of-the-art function calling model with a commercially viable license. Key info and highlights
+💡 The model is also hosted on the [Fireworks](https://fireworks.ai/models/fireworks/firefunction-v1) platform. Offered for free during a limited beta period
+⭐️ Near GPT-4 level quality for real-world use cases of structured information generation and routing decision-making
+💨 Blazing fast speed. Inference speeds are roughly 4x that of GPT-4 when using FireFunction hosted on the Fireworks platform
+🔄 Support for "any" paramter in tool_choice. Firefunction is the only model that we're aware that supports an option for the model to always choose a function - particularly helpful for routing use cases
+✅ The model is also API compatible with [OpenAI function calling](https://platform.openai.com/docs/guides/function-calling).
 ```sh
 OPENAI_API_BASE=https://api.fireworks.ai/inference/v1
 OPENAI_API_KEY=<YOUR_FIREWORKS_API_KEY>
 # Intended Use and Limitations
 ### Primary Use
 Although the model was trained on a variety of tasks, it performs best on:
  * single-turn request routing to a function picked from a pool of up to 20 function specs.