Minueza-32M-UltraChat: A chat model with 32 million parameters

• Base model: Felladrin/Minueza-32M-Base
• Dataset: [ChatML] HuggingFaceH4/ultrachat_200k
• License: Apache License 2.0
• Availability in other ML formats:
  • GGUF: Felladrin/gguf-Minueza-32M-UltraChat
  • ONNX: Felladrin/onnx-Minueza-32M-UltraChat

Recommended Prompt Format

<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{user_message}<|im_end|>
<|im_start|>assistant

Recommended Inference Parameters

do_sample: true
temperature: 0.65
top_p: 0.55
top_k: 35
repetition_penalty: 1.176

Usage Example

from transformers import pipeline

generate = pipeline("text-generation", "Felladrin/Minueza-32M-UltraChat")

messages = [
    {
        "role": "system",
        "content": "You are a highly knowledgeable and friendly assistant. Your goal is to understand and respond to user inquiries with clarity. Your interactions are always respectful, helpful, and focused on delivering the most accurate information to the user.",
    },
    {
        "role": "user",
        "content": "Hey! Got a question for you!",
    },
    {
        "role": "assistant",
        "content": "Sure! What's it?",
    },
    {
        "role": "user",
        "content": "What are some potential applications for quantum computing?",
    },
]

prompt = generate.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

output = generate(
    prompt,
    max_new_tokens=256,
    do_sample=True,
    temperature=0.65,
    top_k=35,
    top_p=0.55,
    repetition_penalty=1.176,
)

print(output[0]["generated_text"])

How it was trained

This model was trained with SFTTrainer using the following settings:

Hyperparameter	Value
Learning rate	2e-5
Total train batch size	16
Max. sequence length	2048
Weight decay	0
Warmup ratio	0.1
Optimizer	Adam with betas=(0.9,0.999) and epsilon=1e-08
Scheduler	cosine
Seed	42

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	28.97
AI2 Reasoning Challenge (25-Shot)	21.08
HellaSwag (10-Shot)	26.95
MMLU (5-Shot)	26.08
TruthfulQA (0-shot)	47.70
Winogrande (5-shot)	51.78
GSM8k (5-shot)	0.23