{
  "peft_type": "LORA",
  "r": 64,
  "lora_alpha": 128,
  "use_rslora": true,
  "target_modules": ["q_proj", "k_proj", "v_proj", "o_proj", "gate_up_proj"],
  "lora_dropout": 0.0
}

Usage

python

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

tokenizer = AutoTokenizer.from_pretrained("arealicehole/trinity-nano-hermes-qlora")
model = AutoModelForCausalLM.from_pretrained(
    "arealicehole/trinity-nano-hermes-qlora",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

messages = [{"role": "user", "content": "What's 84 * 3 / 2?"}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda")
outputs = model.generate(inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0]))

Merge with Base (Transformers)

python

from transformers import AutoModelForCausalLM
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained(
    "arcee-ai/Trinity-Nano-Preview",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
model = PeftModel.from_pretrained(base, "arealicehole/trinity-nano-hermes-qlora")
merged = model.merge_and_unload()

Model Sources

Base Model: arcee-ai/Trinity-Nano-Preview
Training Dataset: NousResearch/hermes-function-calling-v1

Cite

bibtex

@misc{trinity-nano-hermes-qlora,
  author = {arealicehole},
  title = {Trinity Nano Hermes QLoRA},
  year = {2026},
  url = {https://huggingface.co/arealicehole/trinity-nano-hermes-qlora}
}