Reasoning-Llama-1b-v0.1/README.md

base_model, datasets, language, license, tags

base_model	datasets	language	license	tags
meta-llama/Llama-3.2-1B-Instruct	KingNish/reasoning-base-20k	en	llama3.2	text-generation-inference transformers unsloth llama trl sft reasoning llama-3

Model Dexcription

It's First iteration of this model. For testing purpose its just trained on 10k rows. It performed very well than expected. It do first reasoning and than generate response on based on it but it do like o1. It do reasoning separately (Just like o1), no tags (like reflection). Below is inference code.

from transformers import AutoModelForCausalLM, AutoTokenizer

MAX_REASONING_TOKENS = 1024
MAX_RESPONSE_TOKENS = 512

model_name = "KingNish/Reasoning-Llama-1b-v0.1"

model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "Which is greater 9.9 or 9.11 ??"
messages = [
    {"role": "user", "content": prompt}
]

# Generate reasoning
reasoning_template = tokenizer.apply_chat_template(messages, tokenize=False, add_reasoning_prompt=True)
reasoning_inputs = tokenizer(reasoning_template, return_tensors="pt").to(model.device)
reasoning_ids = model.generate(**reasoning_inputs, max_new_tokens=MAX_REASONING_TOKENS)
reasoning_output = tokenizer.decode(reasoning_ids[0, reasoning_inputs.input_ids.shape[1]:], skip_special_tokens=True)

# print("REASONING: " + reasoning_output)

# Generate answer
messages.append({"role": "reasoning", "content": reasoning_output})
response_template = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
response_inputs = tokenizer(response_template, return_tensors="pt").to(model.device)
response_ids = model.generate(**response_inputs, max_new_tokens=MAX_RESPONSE_TOKENS)
response_output = tokenizer.decode(response_ids[0, response_inputs.input_ids.shape[1]:], skip_special_tokens=True)

print("ANSWER: " + response_output)

• Trained by: Nishith Jain
• License: llama3.2
• Finetuned from model : meta-llama/Llama-3.2-1B-Instruct
• Dataset used : KingNish/reasoning-base-20k

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.