Instructions to use cisimcik/cisimcik-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use cisimcik/cisimcik-base with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="cisimcik/cisimcik-base") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("cisimcik/cisimcik-base") model = AutoModelForCausalLM.from_pretrained("cisimcik/cisimcik-base", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use cisimcik/cisimcik-base with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "cisimcik/cisimcik-base" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "cisimcik/cisimcik-base", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/cisimcik/cisimcik-base
- SGLang
How to use cisimcik/cisimcik-base 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 "cisimcik/cisimcik-base" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "cisimcik/cisimcik-base", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'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 "cisimcik/cisimcik-base" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "cisimcik/cisimcik-base", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use cisimcik/cisimcik-base with Docker Model Runner:
docker model run hf.co/cisimcik/cisimcik-base
cisimcik-base
cisimcik-base, Tรผrkรงe รถncelikli bir temel (base) dil modelidir (base model). Qwen/Qwen3.5-4B-Base modelinin,
รงoฤunlukla Tรผrkรงe metinle devam รถn eฤitiminden (continued pre-training) eฤitilmiล halidir.
- 4,2 milyar parametre. Qwen3.5 hibrit mimarisi (Gated DeltaNet + full attention), yalnฤฑzca metin.
- 10B token training. %91,5 Tรผrkรงe, %8,5 replay (ฤฐngilizce, kod, matematik, 23 baลka dil).
- Apache 2.0 lisansฤฑyla aรงฤฑk kaynak olarak yayฤฑnlandฤฑ.
Hฤฑzlฤฑ baลlangฤฑรง
pip install "transformers>=5.17" torch
# isteฤe baฤlฤฑ, NVIDIA GPU'larda daha hฤฑzlฤฑ Gated DeltaNet kernel'leri:
pip install flash-linear-attention
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("cisimcik/cisimcik-base")
model = AutoModelForCausalLM.from_pretrained("cisimcik/cisimcik-base", dtype=torch.bfloat16)
model = model.to("cuda" if torch.cuda.is_available() else "cpu").eval()
inputs = tok("Kapadokya'daki peri bacalarฤฑ,", return_tensors="pt").to(model.device)
out = model.generate(**inputs, max_new_tokens=150, do_sample=True, temperature=0.7, top_p=0.9)
print(tok.decode(out[0], skip_special_tokens=True))
Bu bir temel modeldir. Sadece verilen prompt'u tamamlar, sohbet etmez. Sohbet iรงin cisimcik-4b kullanฤฑn.
Donanฤฑm. Aฤฤฑrlฤฑklar bf16'da yaklaลฤฑk 8,4 GB: 12 GB veya รผzeri belleฤe sahip herhangi bir GPU ya da 16 GB RAM'li
bir CPU yeterlidir. flash-linear-attention kurulu deฤilse Gated DeltaNet katmanlarฤฑ transformers'ฤฑn PyTorch
uygulamasฤฑnฤฑ kullanฤฑr. Bu doฤru sonuรง verir ama uzun girdilerde daha yavaลtฤฑr.
Gรถrev kฤฑyaslamalarฤฑna bu model รผzerine eฤitilen modellerden ulaลฤฑlabilir: cisimcik-4b (CETVEL, OpenLLM Turkish Leaderboard) ve tau-4b.
Mimari
| Parรงa | Ayrฤฑntฤฑ |
|---|---|
| Model | Qwen3.5-4B metin modeli (Qwen3_5ForCausalLM): 32 katman, hidden size 2560, MLP 9216. Her 4. katman full attention (16 query / 4 KV head, head dim 256), diฤer 24'รผ Gated DeltaNet linear attention. |
| Dictionary | 248.320, Qwen3.5 tokenizer'ฤฑ deฤiลtirilmeden. Tied embedding'ler. |
| Temel modelden รงฤฑkarฤฑlanlar | gรถrรผntรผ kodlayฤฑcฤฑ (vision encoder) ve รงoklu token tahmini (MTP) katmanฤฑ |
| Parametre | 4,21 milyar |
| Context | 16.384 token'a kadar dizilerle eฤitildi. Pozisyonlar 262.144'e kadar gider (temel modelden) |
Dosyalar
| Dosya | ฤฐรงerik |
|---|---|
model-0000*-of-00002.safetensors |
aฤฤฑrlฤฑklar (bf16) |
config.json |
Qwen3_5ForCausalLM config'i |
tokenizer*.json, vocab.json, merges.txt |
Qwen3.5 tokenizer'ฤฑ, deฤiลtirilmeden |
Lisans
Bu depodaki aฤฤฑrlฤฑklar Apache License 2.0 ile yayฤฑmlanmฤฑลtฤฑr (bkz. LICENSE ve NOTICE). Model, Qwen/Qwen3.5-4B-Base ยฉ 2026 Alibaba Cloud, Apache License 2.0 lisanslฤฑ modelin deฤiลtirilmiล bir sรผrรผmรผdรผr.
Temel model Qwen/Qwen3.5-4B-Base'dir.
Atฤฑf
@misc{cisimcikbase2026,
title = {cisimcik-base: a Turkish-first base language model},
author = {cisimcik},
year = {2026},
url = {https://huggingface.co/cisimcik/cisimcik-base}
}
- Downloads last month
- 251