Modelpublic

spanthee/gemma4-e4b-macos-clawdia-toolcalling-lora

sourceHugging Faceotherupdated 3mo agoView on Hugging Face

0likes7downloads

Model Card

Gemma 4 E4B macOS Clawdia Tool Calling LoRA

This is a Transformers/PEFT LoRA adapter fine-tuned from google/gemma-4-E4B-it for macOS, Clawdia, computer-use-style tool calling, and receipt/grocery assistant workflows.

It is adapter-only, not a merged full checkpoint. Load it with the base model through peft, or merge it into the base model if you need a standalone Transformers artifact.

Quick start

python

import torch
from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer

model_id = "spanthee/gemma4-e4b-macos-clawdia-toolcalling-lora"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoPeftModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True,
)

messages = [
    {
        "role": "user",
        "content": "Given these available tools, extract this receipt into grocery categories.",
    }
]
prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=256, do_sample=False)
print(tokenizer.decode(output[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))

Training procedure

This model was trained with SFT.

Framework versions

—PEFT 0.19.1
—TRL: 1.6.0
—Transformers: 5.13.0.dev0
—Pytorch: 2.7.0
—Datasets: 5.0.0
—Tokenizers: 0.22.2

Citations

Cite TRL as:

bibtex

@software{vonwerra2020trl,
  title   = {{TRL: Transformers Reinforcement Learning}},
  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
  license = {Apache-2.0},
  url     = {https://github.com/huggingface/trl},
  year    = {2020}
}