CreitinGameplays/Llama-3.1-8b-reasoning-test-Q4_K_M-GGUF

This model was converted to GGUF format from CreitinGameplays/Llama-3.1-8b-reasoning-test using llama.cpp via the ggml.ai's GGUF-my-repo space. Refer to the original model card for more details on the model.

Use with llama.cpp

Install llama.cpp through brew (works on Mac and Linux)

brew install llama.cpp

Invoke the llama.cpp server or the CLI.

CLI:

llama-cli --hf-repo CreitinGameplays/Llama-3.1-8b-reasoning-test-Q4_K_M-GGUF --hf-file llama-3.1-8b-reasoning-test-q4_k_m.gguf -p "The meaning to life and the universe is"

Server:

llama-server --hf-repo CreitinGameplays/Llama-3.1-8b-reasoning-test-Q4_K_M-GGUF --hf-file llama-3.1-8b-reasoning-test-q4_k_m.gguf -c 2048

Note: You can also use this checkpoint directly through the usage steps listed in the Llama.cpp repo as well.

Step 1: Clone llama.cpp from GitHub.

git clone https://github.com/ggerganov/llama.cpp

Step 2: Move into the llama.cpp folder and build it with LLAMA_CURL=1 flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux).

cd llama.cpp && LLAMA_CURL=1 make

Step 3: Run inference through the main binary.

./llama-cli --hf-repo CreitinGameplays/Llama-3.1-8b-reasoning-test-Q4_K_M-GGUF --hf-file llama-3.1-8b-reasoning-test-q4_k_m.gguf -p "The meaning to life and the universe is"

or

./llama-server --hf-repo CreitinGameplays/Llama-3.1-8b-reasoning-test-Q4_K_M-GGUF --hf-file llama-3.1-8b-reasoning-test-q4_k_m.gguf -c 2048

---

Run the model:

from llama_cpp import Llama

# Load the model (using the full training context for inference)
llm = Llama.from_pretrained(
    repo_id="CreitinGameplays/Llama-3.1-8b-reasoning-test-Q4_K_M-GGUF",
    filename="*.gguf",
    verbose=False,
    n_gpu_layers=0,  # CPU-only; increase if using GPU
    n_batch=512,
    n_ctx=8192,
    n_ctx_per_seq=8192,
    f16_kv=True
)

# Set up initial chat history with a system prompt.
chat_history = [
    {"role": "system", "content": """
You are a helpful assistant named Llama, made by Meta AI.
Always use your <|reasoning|> and <|end_reasoning|> tokens, without any text formatting, plain text only.
    """}
]

print("Enter 'quit' or 'exit' to stop the conversation.")

while True:
    # Prompt the user for input
    user_input = input("\nUser: ")
    if user_input.lower() in ["quit", "exit"]:
        break

    # Append the new user message to the chat history.
    chat_history.append({"role": "user", "content": user_input})

    # Call the chat completion API in streaming mode with the updated conversation.
    output_stream = llm.create_chat_completion(
        messages=chat_history,
        temperature=0.4,
        top_p=0.95,
        max_tokens=4096,
        stream=True
    )

    collected_reply = ""
    last_finish_reason = None

    # Process each chunk as it arrives.
    print("Assistant: ", end="", flush=True)
    for chunk in output_stream:
        # Each chunk has a 'choices' list; we get the first choice's delta.
        delta = chunk["choices"][0].get("delta", {})
        if "content" in delta:
            text = delta["content"]
            print(text, end="", flush=True)
            collected_reply += text                                                                                                                                                                                                                                               if "finish_reason" in chunk["choices"][0]:                                                                                                                                                                                                                                    last_finish_reason = chunk["choices"][0]["finish_reason"]

    # Add the assistant's reply to the conversation history.
    chat_history.append({"role": "assistant", "content": collected_reply})
                                                                                                                                                                                                                                                                              # Inform the user if generation stopped due to reaching the token limit.
    if last_finish_reason == "length":
        print("\n[Generation stopped: reached max_tokens. Consider increasing max_tokens or continuing the conversation.]")