Field	Value
Source checkpoint	`ermiaazarkhalili/Gemma4-E2B-Function-Calling-xLAM-Unsloth`
Base model	`google/gemma-4-E2B-it`
Dataset	`Salesforce/xlam-function-calling-60k`
Training	N=1 full epoch (7,500 steps, effective batch=8)
Conversion	Unsloth `save_pretrained_gguf` → llama.cpp GGUF
Quantization tool	llama.cpp `llama-quantize`

Available quantizations

File	Size	Notes
`Gemma4-E2B-Function-Calling-xLAM-Unsloth.Q4_K_M.gguf`	recommended	4-bit; best size/quality balance
`Gemma4-E2B-Function-Calling-xLAM-Unsloth.Q5_K_M.gguf`	balanced	5-bit; near-full quality
`Gemma4-E2B-Function-Calling-xLAM-Unsloth.Q8_0.gguf`	largest	8-bit; closest to bf16 source

Recommended default: Q4_K_M (4-bit, K-quant medium). For maximum fidelity, use Q8_0. (Only 3 of 6 quants are available — 6-quant conversion exceeded memory limits on MIG 3g.40gb.)

Usage

llama.cpp

bash

# Text-only
llama-cli -hf ermiaazarkhalili/Gemma4-E2B-Function-Calling-xLAM-Unsloth-GGUF --jinja -p "Find flights from SFO to NYC on December 25th" -n 256

# Interactive chat
llama-cli -hf ermiaazarkhalili/Gemma4-E2B-Function-Calling-xLAM-Unsloth-GGUF --jinja -cnv

Ollama

ollama run hf.co/ermiaazarkhalili/Gemma4-E2B-Function-Calling-xLAM-Unsloth-GGUF:Q4_K_M

llama-cpp-python

python

from llama_cpp import Llama
llm = Llama.from_pretrained(
    repo_id="ermiaazarkhalili/Gemma4-E2B-Function-Calling-xLAM-Unsloth-GGUF",
    filename="*Q4_K_M.gguf",
    n_ctx=2048,
)
out = llm.create_chat_completion(
    messages=[{"role": "user", "content": "Find flights from SFO to NYC on December 25th"}],
    max_tokens=256,
)
print(out["choices"][0]["message"]["content"])

Intended use

For research and non-commercial experimentation only. Outputs should be independently verified before any downstream use.

Limitations

GGUF quantizations have unavoidable quality loss relative to the source bfloat16 checkpoint. Use Q5_K_M or Q8_0 for best fidelity.
Inherits all limitations of the source merged checkpoint (ermiaazarkhalili/Gemma4-E2B-Function-Calling-xLAM-Unsloth).
Limited to the 60 function schemas covered in the training dataset; performance on novel APIs may degrade. Although the base model is a vision-language model (VLM), this fine-tune targets text-only function-calling; vision capabilities are untouched but unevaluated. This repo ships 3 of 6 planned quants (Q4_K_M, Q5_K_M, Q8_0). The repo also includes a BF16-mmproj.gguf sidecar for the vision projection; use --mmproj with llama-cli if doing multimodal inference, though this fine-tune targets text-only function-calling/reasoning.

Citation

bibtex

@misc{ gemma4_e2b_xlam_unsloth_2026_gguf ,
  author = {Ermia Azarkhalili},
  title = { Gemma4-E2B-xLAM-Unsloth — GGUF quantized },
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/ermiaazarkhalili/Gemma4-E2B-Function-Calling-xLAM-Unsloth-GGUF}}
}

This gemma4 model was trained 2× faster with Unsloth and Hugging Face's TRL library.