Instructions to use ArtyomSubDiv/pythia-14m-dd-rus-draft with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ArtyomSubDiv/pythia-14m-dd-rus-draft with Transformers:

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

pipe = pipeline("text-generation", model="ArtyomSubDiv/pythia-14m-dd-rus-draft")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ArtyomSubDiv/pythia-14m-dd-rus-draft")
model = AutoModelForCausalLM.from_pretrained("ArtyomSubDiv/pythia-14m-dd-rus-draft", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use ArtyomSubDiv/pythia-14m-dd-rus-draft 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 ArtyomSubDiv/pythia-14m-dd-rus-draft:F16
# Run inference directly in the terminal:
llama cli -hf ArtyomSubDiv/pythia-14m-dd-rus-draft:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ArtyomSubDiv/pythia-14m-dd-rus-draft:F16
# Run inference directly in the terminal:
llama cli -hf ArtyomSubDiv/pythia-14m-dd-rus-draft:F16

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 ArtyomSubDiv/pythia-14m-dd-rus-draft:F16
# Run inference directly in the terminal:
./llama-cli -hf ArtyomSubDiv/pythia-14m-dd-rus-draft:F16

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 ArtyomSubDiv/pythia-14m-dd-rus-draft:F16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ArtyomSubDiv/pythia-14m-dd-rus-draft:F16

Use Docker

docker model run hf.co/ArtyomSubDiv/pythia-14m-dd-rus-draft:F16

LM Studio
Jan

vLLM

How to use ArtyomSubDiv/pythia-14m-dd-rus-draft with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ArtyomSubDiv/pythia-14m-dd-rus-draft"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ArtyomSubDiv/pythia-14m-dd-rus-draft",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/ArtyomSubDiv/pythia-14m-dd-rus-draft:F16

SGLang

How to use ArtyomSubDiv/pythia-14m-dd-rus-draft 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 "ArtyomSubDiv/pythia-14m-dd-rus-draft" \
    --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": "ArtyomSubDiv/pythia-14m-dd-rus-draft",
		"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 "ArtyomSubDiv/pythia-14m-dd-rus-draft" \
        --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": "ArtyomSubDiv/pythia-14m-dd-rus-draft",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Ollama
How to use ArtyomSubDiv/pythia-14m-dd-rus-draft with Ollama:
```
ollama run hf.co/ArtyomSubDiv/pythia-14m-dd-rus-draft:F16
```
Unsloth Desktop
Docker Model Runner
How to use ArtyomSubDiv/pythia-14m-dd-rus-draft with Docker Model Runner:
```
docker model run hf.co/ArtyomSubDiv/pythia-14m-dd-rus-draft:F16
```

Lemonade

How to use ArtyomSubDiv/pythia-14m-dd-rus-draft with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ArtyomSubDiv/pythia-14m-dd-rus-draft:F16

Run and chat with the model

lemonade run user.pythia-14m-dd-rus-draft-F16

List all available models

lemonade list

Atomic Chat

К сожалению, это весьма неудачная попытка дообучить базовую Pythia 14M Deduped.

При обучении, mean_token_accuracy и loss достигнув значений около 0.555 и 1.9 практически не менялись на протяжении кучи эпох независимо от того, каким был LR.

Полагаю, без хорошего pretrain на русском языке сделать из неё нормальную Instruct модель на нём же не удастся, так что, на данный момент, она заучила лишь слова и форму построения предложений, но генерит лишь полный рандом, и на конкретные инструкции из датасета даже при предсказуемом сэмплировании даёт другие ответы.

Есть некоторые сомнения, что хоть когда-нибудь её получится обучить, так как уже есть хороший пример от OpenAssistant, когда они воспользовались претрейном Pythia-12B, а потом уже обучили его на своём датасете oasst, и в итоге модель получилась слабой.

Downloads last month: 544

Safetensors

Model size

14.1M params

Tensor type

BF16

Model tree for ArtyomSubDiv/pythia-14m-dd-rus-draft

Base model

EleutherAI/pythia-14m-deduped

Finetuned

(2)

this model

ArtyomSubDiv
/

pythia-14m-dd-rus-draft

Model tree for ArtyomSubDiv/pythia-14m-dd-rus-draft

Dataset used to train ArtyomSubDiv/pythia-14m-dd-rus-draft