Instructions to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Use Docker

docker model run hf.co/junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "junwatu/resep-ID-chat-gemma-4-E4B-it-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "junwatu/resep-ID-chat-gemma-4-E4B-it-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Ollama
How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with Ollama:
```
ollama run hf.co/junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M
```
Unsloth Desktop

How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with Docker Model Runner:
```
docker model run hf.co/junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M
```

Lemonade

How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Run and chat with the model

lemonade run user.resep-ID-chat-gemma-4-E4B-it-gguf-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use junwatu/resep-ID-chat-gemma-4-E4B-it-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "junwatu/resep-ID-chat-gemma-4-E4B-it-gguf:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Resep ID Chat — Gemma 4 E4B (GGUF)

GGUF quantization of junwatu/resep-ID-chat-gemma-4-E4B-it, a full fine-tune of google/gemma-4-E4B-it for Indonesian recipe generation.

This is the portable, runs-on-your-laptop version. Use it with llama.cpp, Ollama, LM Studio, or any GGUF-compatible runtime.

Quants in this repo

File	Approx size	Quality	Recommended for
`resep-ID-chat-gemma-4-E4B-it.Q4_K_M.gguf`	~5–6 GB	~96% of bf16	Default. Runs on a laptop with 12 GB+ RAM.
`resep-ID-chat-gemma-4-E4B-it.Q8_0.gguf`	~8–9 GB	~99.5% of bf16	Near-lossless. For 16+ GB RAM, when you want maximum quality without the full bf16 download.

The Q4_K_M↔Q5_K_M gap is small relative to the embedding overhead from Gemma 4's large vocab, so Q5 was skipped — jump to Q8_0 if you want noticeably better quality.

⚠️ Critical inference setting (different from the e2b release)

Use a standard repetition penalty of 1.10, and DO NOT enable an n-gram or DRY repetition blocker.

The e2b release of this fine-tune required no_repeat_ngram_size=6 (or llama.cpp's DRY sampler with allowed_length=6) to break bumbu-list mode-collapse. On this larger e4b model the same setting causes artifacts (Unicode fractions, invented unit abbreviations, morphology variants). A 50-sample eval validated that disabling the n-gram blocker and slightly bumping the repetition penalty removes every artifact pattern.

Translation per runtime:

Runtime	E2B release setting	E4B (this repo) setting
HuggingFace `transformers`	`no_repeat_ngram_size=6`, `repetition_penalty=1.05`	`no_repeat_ngram_size=0`, `repetition_penalty=1.10`
llama.cpp	DRY sampler with `--dry-allowed-length 6`	No DRY sampler. Just `--repeat-penalty 1.10`
LM Studio	Enable "DRY" with allowed-length 6	Disable DRY. Set Repeat Penalty = 1.10
Ollama	`repeat_penalty 1.05`	`repeat_penalty 1.10`

Quick start

llama.cpp (recommended)

./llama-cli \
  -m resep-ID-chat-gemma-4-E4B-it.Q4_K_M.gguf \
  --repeat-penalty 1.10 \
  -p "Tulis resep masakan Indonesia berjudul: \"Tumis Kangkung Tempe\".
Format jawaban:
Bahan:
- (daftar bahan, satu per baris)

Langkah:
1. (langkah pertama)
2. (langkah kedua)
...
Gunakan Bahasa Indonesia."

For server mode (OpenAI-compatible API):

./llama-server \
  -m resep-ID-chat-gemma-4-E4B-it.Q4_K_M.gguf \
  --repeat-penalty 1.10 \
  -c 4096 --host 127.0.0.1 --port 8080

LM Studio

Search for junwatu/resep-ID-chat-gemma-4-E4B-it-gguf in the model browser.
Download Q4_K_M.
In Inference Settings:
- Set Repeat Penalty to 1.10.
- Make sure DRY sampler is disabled (it's enabled by default if you copied settings from the e2b release).

Ollama

cat > Modelfile <<EOF
FROM ./resep-ID-chat-gemma-4-E4B-it.Q4_K_M.gguf
PARAMETER repeat_penalty 1.10
PARAMETER temperature 0.0
EOF
ollama create resep-id-e4b -f Modelfile
ollama run resep-id-e4b "Tulis resep masakan Indonesia berjudul: \"Tumis Kangkung Tempe\"..."

Recommended system prompt (domain guardrail)

To make the model politely refuse off-topic queries (math, code, news, etc.), prepend this to the user message:

Kamu adalah asisten resep masakan Indonesia. Tugasmu hanya membuat resep,
menjelaskan bahan masakan, dan memberi tips memasak. Jika pengguna bertanya
tentang topik lain, tolak dengan sopan dalam Bahasa Indonesia dan ingatkan
kalau kamu hanya bisa bantu soal masak-memasak.

---

Tulis resep masakan Indonesia berjudul: "<dish title>".
...

Effectiveness: ~70–80% of off-topic queries get politely declined. Without the guardrail, the model will hallucinate "recipes" for non-recipe titles.

What this model does

Give it an Indonesian recipe title or a list of ingredients, get back a structured Bahan: ... Langkah: ... recipe in natural Bahasa Indonesia. Trained on ~~183K real home-cook recipes (~~3× the e2b release dataset).

See the base model card for the full description, comparison to the e2b release, training summary, and known limitations.

Quality vs the bf16 original

Q4_K_M typically loses ~3-5% quality vs the full-precision model. For recipe generation that's largely imperceptible — same dish identity, same format, same step coherence. The full bf16 original lives at junwatu/resep-ID-gemma-4-E4B-it if you need maximum quality.

License

Inherits the Gemma Terms of Use from google/gemma-4-E4B-it.

Downloads last month: -

GGUF

Model size

7B params

Architecture

gemma4

Hardware compatibility

4-bit

8-bit

Model tree for junwatu/resep-ID-chat-gemma-4-E4B-it-gguf

Base model

google/gemma-4-E4B

Finetuned

google/gemma-4-E4B-it

Finetuned

junwatu/resep-ID-chat-gemma-4-E4B-it

Quantized

(1)

this model