Instructions to use wyan/usenet-gemma-4-E2B-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use wyan/usenet-gemma-4-E2B-lora with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("google/gemma-4-E2B")
model = PeftModel.from_pretrained(base_model, "wyan/usenet-gemma-4-E2B-lora")

Transformers

How to use wyan/usenet-gemma-4-E2B-lora with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="wyan/usenet-gemma-4-E2B-lora")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("wyan/usenet-gemma-4-E2B-lora", dtype="auto")

llama-cpp-python

How to use wyan/usenet-gemma-4-E2B-lora with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="wyan/usenet-gemma-4-E2B-lora",
	filename="usenet-gemma-v0.01.Q4_K_M.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use wyan/usenet-gemma-4-E2B-lora with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf wyan/usenet-gemma-4-E2B-lora:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf wyan/usenet-gemma-4-E2B-lora:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf wyan/usenet-gemma-4-E2B-lora:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf wyan/usenet-gemma-4-E2B-lora: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 wyan/usenet-gemma-4-E2B-lora:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf wyan/usenet-gemma-4-E2B-lora: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 wyan/usenet-gemma-4-E2B-lora:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf wyan/usenet-gemma-4-E2B-lora:Q4_K_M

Use Docker

docker model run hf.co/wyan/usenet-gemma-4-E2B-lora:Q4_K_M

LM Studio
Jan

vLLM

How to use wyan/usenet-gemma-4-E2B-lora with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "wyan/usenet-gemma-4-E2B-lora"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "wyan/usenet-gemma-4-E2B-lora",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/wyan/usenet-gemma-4-E2B-lora:Q4_K_M

SGLang

How to use wyan/usenet-gemma-4-E2B-lora with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "wyan/usenet-gemma-4-E2B-lora" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "wyan/usenet-gemma-4-E2B-lora",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "wyan/usenet-gemma-4-E2B-lora" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "wyan/usenet-gemma-4-E2B-lora",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Ollama
How to use wyan/usenet-gemma-4-E2B-lora with Ollama:
```
ollama run hf.co/wyan/usenet-gemma-4-E2B-lora:Q4_K_M
```

Unsloth Studio new

How to use wyan/usenet-gemma-4-E2B-lora 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 wyan/usenet-gemma-4-E2B-lora 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 wyan/usenet-gemma-4-E2B-lora to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for wyan/usenet-gemma-4-E2B-lora to start chatting

Docker Model Runner
How to use wyan/usenet-gemma-4-E2B-lora with Docker Model Runner:
```
docker model run hf.co/wyan/usenet-gemma-4-E2B-lora:Q4_K_M
```

Lemonade

How to use wyan/usenet-gemma-4-E2B-lora with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull wyan/usenet-gemma-4-E2B-lora:Q4_K_M

Run and chat with the model

lemonade run user.usenet-gemma-4-E2B-lora-Q4_K_M

List all available models

lemonade list

Notes from the dataset creator - tips for better fine-tuning

by OwnedByDanes - opened 10 days ago

Discussion

OwnedByDanes

10 days ago

This is fantastic! Really glad to see someone actually shipping something with the corpus. A few thoughts that might help with the "works terribly" problem:

For a Usenet-flavored model, the comp.* and sci.* samples will give you much cleaner signal than alt.* — the alt hierarchy has a lot of noise and the writing style is all over the place. If you haven't already, try fine-tuning on just the comp-sample or sci-sample files and see if the output quality improves noticeably.

Also worth filtering by date if you can... the pre-1995 posts have a very distinct voice (longer, more technical, less conversational) compared to the late 90s material. Depending on which era you're going for, mixing those can confuse the model.

Keep iterating! The fact that you got a quantized GGUF running on ollama with downloads already is great progress. Happy to answer any questions about the data structure or content if it would help.

wyan

Owner 1 day ago

Thank you for your encouragement!

I must admit I don't really know what I'm doing, this is my first time tuning models or running any ML at all, and I'm learning as I go. I'll try your suggestions. One thing I still don't quite get is what's the "correct" format to give the data to feed the tuning process. This attempt uses the subject line as a sort of question and the body as an answer, which I'm guessing makes it more like reinforcement learning and less like the fine tuning we want here. Need to read up on this.

I'm hoping to find some time to give it another go soon, it's pretty impressive in any case to be able to play with this sort of algorithm, at this size, as a hobby :)

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment