function_calling_llama3_8b_instruct

Tool Information

Define the tools and their functionalities as a list of dictionaries.

tools_info = [
  {
    "name": "cancel_reservation",
    "description": "cancel a reservation",
    "parameters": {
      "type": "object",
      "properties": {
        "reservation_number": {
          "type": "integer",
          "description": "Reservation number"
        }
      },
      "required": ["reservation_number"]
    }
  },
  {
    "name": "get_reservations",
    "description": "get reservation numbers",
    "parameters": {
      "type": "object",
      "properties": {
        "user_id": {
          "type": "integer",
          "description": "User id"
        }
      },
      "required": [
        "user_id"
      ]
    }
  },
]

System Initialization

Initialize the system's interactive capabilities using the defined tools.

system = f"You are a helpful assistant with access to the following functions: \n {json.dumps(tools_info, indent=2)}."

Conversation Flow

Simulate a conversation flow where the user requests to cancel a reservation.

messages = [
    {"role": "system", "content": system},
    {"role": "user", "content": "Help me to cancel a reservation"},
    {"role": "assistant", "content": "I can help with that. Could you please provide me with the reservation number?"},
    {"role": "user", "content": "the reservation number is 1011"}
]

Or the user requests to display its reservations, note the use of "tool" role.

messages=[
    {"role":"system","content": system},
    {"role": "user","content": "Help me to find my reservations, my user id is 110"},
    {"role": "assistant","content":'<func_call> {"name": "get_reservations", "arguments": {"user_id": 110}}'},
    {"role": "tool","content":'["AB001","CD002","GG100"]'}
]

Model Loading

Load the causal language model and tokenizer.

model_id = "caldana/function_calling_llama3_8b_instruct"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

Generating Response

Generate a response from the model based on the conversation context.

input_ids = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt"
).to(model.device)

terminators = [
    tokenizer.eos_token_id,
    tokenizer.convert_tokens_to_ids("")
]

outputs = model.generate(
    input_ids,
    max_new_tokens=256,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.9,
)

response = outputs[0][input_ids.shape[-1]:]
print(tokenizer.decode(response, skip_special_tokens=True))