metadata

language:
  - en
license: llama3.2
tags:
  - text-generation-inference
  - transformers
  - llama
  - trl
  - sft
  - reasoning
  - llama-3
base_model: lunahr/Hermes-3-Llama-3.2-3B-abliterated
datasets:
  - KingNish/reasoning-base-20k
  - lunahr/thea-name-overrides

Note

Thea v2 has not been renamed on a weight level. It will use default responses instead.

Model Description

An uncensored reasoning Llama 3.2 3B model trained on reasoning data.

This is the 2nd revision of Thea, based on a better base model, and with twice the reasoning data.

It has been trained using improved training code, and gives an improved performance. Here is what inference code you should use:

from transformers import AutoModelForCausalLM, AutoTokenizer

MAX_REASONING_TOKENS = 1024
MAX_RESPONSE_TOKENS = 512

model_name = "lunahr/thea-v2-3b-50r"

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: Piotr Zalewski
License: llama3.2
Finetuned from model: lunahr/Hermes-3-Llama-3.2-3B-abliterated*
Dataset used: KingNish/reasoning-base-20k

This Llama model was trained faster than Unsloth using custom training code.

Visit https://www.kaggle.com/code/piotr25691/distributed-llama-training-with-2xt4 to find out how you can finetune your models using BOTH of the Kaggle provided GPUs.

*Created from https://huggingface.co/NousResearch/Hermes-3-Llama-3.2-3B using a custom abliterator.