Instructions to use coderian/TanAi-turkish-29M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use coderian/TanAi-turkish-29M with Transformers:

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

pipe = pipeline("text-generation", model="coderian/TanAi-turkish-29M", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("coderian/TanAi-turkish-29M", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use coderian/TanAi-turkish-29M with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "coderian/TanAi-turkish-29M"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "coderian/TanAi-turkish-29M",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/coderian/TanAi-turkish-29M

SGLang

How to use coderian/TanAi-turkish-29M 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 "coderian/TanAi-turkish-29M" \
    --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": "coderian/TanAi-turkish-29M",
		"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 "coderian/TanAi-turkish-29M" \
        --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": "coderian/TanAi-turkish-29M",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use coderian/TanAi-turkish-29M with Docker Model Runner:
```
docker model run hf.co/coderian/TanAi-turkish-29M
```

TanAi-turkish-29M

TanAI mimarisiyle sıfırdan eğitilmiş, 28.9M parametreli deneysel bir Türkçe dil modeli. Ethosoft/Turkish_corpus veri setinin streaming ile okunan ilk 30.000 örneği üzerinde 2 epoch eğitilmiştir.

Model Detayları

Özellik	Değer
Parametre sayısı	28.9M (28.923.904)
Mimari	TanAI (decoder-only Transformer, GPT benzeri)
Katman sayısı	4
Embedding boyutu	256
Bağlam uzunluğu	128 token
Vocab boyutu	50.257 (GPT-2 tokenizer)
Eğitim verisi	Ethosoft/Turkish_corpus, ilk 30.000 örnek
Epoch	2
Optimizer	AdamW, lr 3e-4
Batch / blok boyutu	16 / 128

Kullanım

Model özel bir mimari kullandığı için trust_remote_code=True gerekir. İlk kullanımda model kodu ve ağırlıklar otomatik olarak indirilir.

from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "coderian/TanAi-turkish-29M"

tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(repo_id, trust_remote_code=True)

inputs = tokenizer("Türkiye'nin en kalabalık şehri", return_tensors="pt")
outputs = model.generate(
    **inputs,
    max_new_tokens=50,
    do_sample=True,
    top_k=50,
    temperature=0.8,
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

pipeline ile:

from transformers import pipeline

generator = pipeline("text-generation", model=repo_id, trust_remote_code=True)
print(generator("Bugün hava", max_new_tokens=30)[0]["generated_text"])

İndirme (isteğe bağlı, from_pretrained modeli zaten otomatik indirir):

huggingface-cli download coderian/TanAi-turkish-29M

Eğitim

Veri, streaming=True ile okundu; yalnızca ilk 30.000 örnek kullanıldı.
Tokenizer: GPT2TokenizerFast; belgeler arasına eos token'ı eklendi.
Kayıp: token bazlı cross-entropy (causal language modeling).
Hiperparametreler: 2 epoch, batch 16, blok 128, AdamW, lr 3e-4.

Sınırlamalar

Küçük ölçekli ve deneysel bir modeldir; çıktılar tekrarlı veya anlamsız olabilir.
Talimat takibi için eğitilmemiştir (instruction tuning yoktur).
Bağlam uzunluğu 128 tokendır; prompt + üretilen token toplamı 128'i geçemez.
Yalnızca Türkçe metin için eğitilmiştir.

Downloads last month: 286

Safetensors

Model size

28.9M params

Tensor type

F32

coderian
/

TanAi-turkish-29M

TanAi-turkish-29M

Model Detayları

Kullanım

Eğitim

Sınırlamalar

Dataset used to train coderian/TanAi-turkish-29M