Instructions to use Bogula/pinktilde30b_64k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Bogula/pinktilde30b_64k with Transformers:

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

pipe = pipeline("text-generation", model="Bogula/pinktilde30b_64k")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Bogula/pinktilde30b_64k")
model = AutoModelForCausalLM.from_pretrained("Bogula/pinktilde30b_64k", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Bogula/pinktilde30b_64k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Bogula/pinktilde30b_64k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Bogula/pinktilde30b_64k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Bogula/pinktilde30b_64k

SGLang

How to use Bogula/pinktilde30b_64k 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 "Bogula/pinktilde30b_64k" \
    --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": "Bogula/pinktilde30b_64k",
		"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 "Bogula/pinktilde30b_64k" \
        --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": "Bogula/pinktilde30b_64k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Bogula/pinktilde30b_64k with Docker Model Runner:
```
docker model run hf.co/Bogula/pinktilde30b_64k
```

TildeOpen-30B · PINKpro (DE, 64k)

LoRA-SFT (bf16, keine Quantisierung) auf Basis von TildeAI/TildeOpen-30b-64k — ein deutschsprachiger, RAG-orientierter Lernassistent (Fuehrung / Vertrieb / Paedagogik) mit Kontextfenster bis 64k.

Training (zweiphasig)

Phase A (kurz/mittel, 8k): Nemotron-Instruction-Following-Chat (reasoning_off) + deutsche Wikipedia-Zusammenfassungen (Wirtschaft/Psychologie/Paedagogik) + Firmen-Single-Turn.
Phase B (Long-Context, bis 64k): Firmen-Multiturn + RAG-Dokumente, mit Distraktoren aufgefuellt.
Nur Anoynme Daten ohne Personen-/Firmenbezug.

Chat-Format

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained(REPO)
model = AutoModelForCausalLM.from_pretrained(REPO, dtype="bfloat16", device_map="auto")
msgs = [{"role":"system","content":"... ## Kontextdokumente ... [DOK 1] ..."},
        {"role":"user","content":"Deine Frage"}]
ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt").to(model.device)
print(tok.decode(model.generate(ids, max_new_tokens=600)[0][ids.shape[1]:], skip_special_tokens=True))

Attribution / Lizenz

Basismodell: TildeAI/TildeOpen-30b-64k.
Trainingsdaten enthalten NVIDIA Nemotron (Lizenz ODC-By) → Attribution erforderlich.

Limitierungen

Ausgaben koennen fehlerhaft sein; nicht fuer Rechts-/Finanz-/Medizin-Entscheidungen ungeprueft nutzen.

Downloads last month: 353

Safetensors

Model size

31B params

Tensor type

BF16

Model tree for Bogula/pinktilde30b_64k

Base model

TildeAI/TildeOpen-30b-64k

Adapter

(2)

this model