Instructions to use voidash/gemma-helpdesk-v4b-e2b-seed42 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries
PEFT
How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with PEFT:
```
Task type is invalid.
```

How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="voidash/gemma-helpdesk-v4b-e2b-seed42",
	filename="gguf/gemma-helpdesk-v4b-step600-e2b-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf voidash/gemma-helpdesk-v4b-e2b-seed42: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 voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf voidash/gemma-helpdesk-v4b-e2b-seed42: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 voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M

Use Docker

docker model run hf.co/voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M

LM Studio
Jan

vLLM

How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "voidash/gemma-helpdesk-v4b-e2b-seed42"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "voidash/gemma-helpdesk-v4b-e2b-seed42",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M

Ollama
How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with Ollama:
```
ollama run hf.co/voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M
```

Unsloth Studio

How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for voidash/gemma-helpdesk-v4b-e2b-seed42 to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for voidash/gemma-helpdesk-v4b-e2b-seed42 to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for voidash/gemma-helpdesk-v4b-e2b-seed42 to start chatting

How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf voidash/gemma-helpdesk-v4b-e2b-seed42: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 voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M

Run Hermes

hermes

Docker Model Runner
How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with Docker Model Runner:
```
docker model run hf.co/voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M
```

Lemonade

How to use voidash/gemma-helpdesk-v4b-e2b-seed42 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull voidash/gemma-helpdesk-v4b-e2b-seed42:Q4_K_M

Run and chat with the model

lemonade run user.gemma-helpdesk-v4b-e2b-seed42-Q4_K_M

List all available models

lemonade list

GovSpeak / PreVillage Gemma E2B v4b

Best E2B release candidate for the GovSpeak / PreVillage service navigator, with a staged llama.cpp GGUF build for edge and kiosk testing.

This is still a retrieval-bound model. Do not treat it as a standalone source of government facts. Fees, contacts, office holders, URLs, required documents, and office-specific details should come from retrieval, structured source packs, deterministic extraction, officer interviews, WhatsApp/citizen reports, and human review.

Recommended E2B Artifact

Use the step600 checkpoint for edge demos:

step600/
gguf/gemma-helpdesk-v4b-step600-e2b-Q4_K_M.gguf

Why step600: it is the best balanced E2B checkpoint from the v4b evals. The best/ checkpoint had stronger refusal correctness but poor Roman-Nepali behavior. Step600 preserved Roman-Nepali behavior and avoided wrong refusals on the grounded gold set.

Step600 Eval

Eval path:

eval/reports/sft_v4b_step600_baseline300_full_eval/

Signal	Result
Grounded items	73
chrF	22.81
URL recall	0.75
Wrong refusals	0/73 = 0.0%
Refusal correctness	83.5%
Belebele Nepali	58.0%
GSM8K-en	53.3%
Roman-Nepali degeneration	0/10

Known limitation: refusal correctness is below the 90% target. Put the model behind resolver and retrieval gates, and prefer deterministic refusal/follow-up logic where source coverage is missing.

GGUF Smoke

The Q4_K_M GGUF loaded in llama.cpp:

model: gguf/gemma-helpdesk-v4b-step600-e2b-Q4_K_M.gguf
prompt throughput: 449.3 t/s
generation throughput: 132.0 t/s

That smoke was on local hardware, not Raspberry Pi. Pi evidence should use the separate Pi runbook/benchmark numbers.

Usage

hf download voidash/gemma-helpdesk-v4b-e2b-seed42 \
  gguf/gemma-helpdesk-v4b-step600-e2b-Q4_K_M.gguf \
  --local-dir ./models

llama-cli \
  -m ./models/gguf/gemma-helpdesk-v4b-step600-e2b-Q4_K_M.gguf \
  --jinja \
  -sys "You are GovSpeak. Answer only from provided sources. Ask a compact follow-up if the service case is ambiguous." \
  -p "Question: mero nagarikta harayo, ke garne?"

Relation To E4B

E2B is the edge/local lane. The stronger planner/composer candidate is the E4B v6.4 adapter:

voidash/gemma-helpdesk-v6-4-e4b-g6e-qlora-seed42

Use E4B where a helpdesk PC or server can run the answer layer. Use E2B/GGUF where a low-cost local device is the deployment constraint.

Downloads last month: 25

GGUF

Model size

5B params

Architecture

gemma4

Hardware compatibility

4-bit

Model tree for voidash/gemma-helpdesk-v4b-e2b-seed42

Base model

google/gemma-4-E2B

Finetuned

google/gemma-4-E2B-it

Adapter

(94)

this model

Collections including voidash/gemma-helpdesk-v4b-e2b-seed42

GovSpeak / PreVillage best models

Collection

Best E4B/E2B Gemma, GGUF, ASR, TTS, and eval assets. • 10 items • Updated 19 days ago

GovSpeak / PreVillage model stack

Collection

Best Gemma, GGUF, Nepali ASR, and Nepali TTS assets for GovSpeak / PreVillage. • 6 items • Updated 19 days ago