Text Generation
Transformers
Safetensors
Turkish
qwen3
turkish
türkçe
continued-pretraining
turkish-llm
türkçe-dil-modeli
conversational
Eval Results (legacy)
text-generation-inference
Instructions to use ecloudtech/Erk-32B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ecloudtech/Erk-32B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="ecloudtech/Erk-32B") 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("ecloudtech/Erk-32B") model = AutoModelForCausalLM.from_pretrained("ecloudtech/Erk-32B", 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 ecloudtech/Erk-32B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "ecloudtech/Erk-32B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "ecloudtech/Erk-32B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/ecloudtech/Erk-32B
- SGLang
How to use ecloudtech/Erk-32B 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 "ecloudtech/Erk-32B" \ --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": "ecloudtech/Erk-32B", "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 "ecloudtech/Erk-32B" \ --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": "ecloudtech/Erk-32B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use ecloudtech/Erk-32B with Docker Model Runner:
docker model run hf.co/ecloudtech/Erk-32B
egitim/: v2 betikleri ve veri ozetleri
Browse files- egitim/dogal_secim.py +43 -0
egitim/dogal_secim.py
ADDED
|
@@ -0,0 +1,43 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# -*- coding: utf-8 -*-
|
| 2 |
+
"""Dogal Turkce metin secimi (FineWeb-2 tur_Latn, parca 100-259: Erk'in daha once GORMEDIGI parcalar).
|
| 3 |
+
GENEL KURAL, KIYASA BAKILMAZ: kiyas maddesi, sablonu, govde listesi, TDK sozlugu KULLANILMAZ.
|
| 4 |
+
Iki havuz:
|
| 5 |
+
A) rastgele paragraflar
|
| 6 |
+
B) kesme isaretli cekimli soz (Istanbul'a, TCDD'yi, 1990'da) yogun paragraflar: >=2 / 100 kelime
|
| 7 |
+
— genel bir yazim olcutu; ozel ad / kisaltma / sayi cekimi dogal metinde en cok burada gorulur.
|
| 8 |
+
Karisim: %70 A + %30 B. Hijyen: sozluk-maddesi gorunumlu satirlar ("kalp, -bi" bicimi) ATILIR ki
|
| 9 |
+
cevap anahtari benzeri metin egitime girmesin. Cikti: dogal_metin.txt (paragraf/satir)."""
|
| 10 |
+
import glob, random, re, sys, unicodedata
|
| 11 |
+
import pyarrow.parquet as pq
|
| 12 |
+
HEDEF_MB = float(sys.argv[1]) if len(sys.argv) > 1 else 600
|
| 13 |
+
KOK = "/ari/users/ytanis/.cache/nanochat/base_data_tr"
|
| 14 |
+
rng = random.Random(20260926)
|
| 15 |
+
KESME = re.compile(r"\b[\wÇĞİÖŞÜçğıöşü]+['’´][a-zçğıöşü]{1,6}\b")
|
| 16 |
+
SOZLUK_MADDESI = re.compile(r"\b[a-zçğıöşü]{2,}, -[a-zçğıöşü]{1,3}\b") # "kalp, -bi"
|
| 17 |
+
TR_HARF = re.compile(r"[çğıöşüÇĞİÖŞÜ]")
|
| 18 |
+
def paragraflar(doc):
|
| 19 |
+
for p in re.split(r"\n\s*\n|\n", doc):
|
| 20 |
+
p = unicodedata.normalize("NFC", p.strip())
|
| 21 |
+
if len(p) < 200 or len(p) > 3000: continue
|
| 22 |
+
k = p.split()
|
| 23 |
+
if len(k) < 25: continue
|
| 24 |
+
if not TR_HARF.search(p): continue # Turkce degil / transliterasyon
|
| 25 |
+
if SOZLUK_MADDESI.search(p): continue # hijyen
|
| 26 |
+
if "TurkMorfBench" in p or "turkmorf" in p.lower(): continue
|
| 27 |
+
yield p, len(KESME.findall(p)) * 100.0 / len(k)
|
| 28 |
+
A, B = [], []
|
| 29 |
+
hedef = HEDEF_MB * 1e6; boyutA = boyutB = 0
|
| 30 |
+
dosyalar = sorted(glob.glob(KOK + "/shard_*.parquet"))[100:260]; rng.shuffle(dosyalar)
|
| 31 |
+
for d in dosyalar:
|
| 32 |
+
for satir in pq.read_table(d, columns=["text"]).column("text").to_pylist():
|
| 33 |
+
for p, yog in paragraflar(satir):
|
| 34 |
+
if yog >= 2.0 and boyutB < 0.30 * hedef:
|
| 35 |
+
B.append(p); boyutB += len(p)
|
| 36 |
+
elif rng.random() < 0.15 and boyutA < 0.70 * hedef:
|
| 37 |
+
A.append(p); boyutA += len(p)
|
| 38 |
+
print("%s A %.0f MB B %.0f MB" % (d[-19:], boyutA / 1e6, boyutB / 1e6), flush=True)
|
| 39 |
+
if boyutA >= 0.70 * hedef and boyutB >= 0.30 * hedef: break
|
| 40 |
+
hepsi = A + B; rng.shuffle(hepsi)
|
| 41 |
+
with open("/ari/users/ytanis/dogal_metin.txt", "w", encoding="utf8") as f:
|
| 42 |
+
for p in hepsi: f.write(p.replace("\n", " ") + "\n")
|
| 43 |
+
print("YAZILDI: %d paragraf (A %d, B %d), %.0f MB" % (len(hepsi), len(A), len(B), (boyutA + boyutB) / 1e6))
|