Instructions to use RaspizdAI/tolk-flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RaspizdAI/tolk-flash with Transformers:

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

pipe = pipeline("text-generation", model="RaspizdAI/tolk-flash")
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("RaspizdAI/tolk-flash")
model = AutoModelForCausalLM.from_pretrained("RaspizdAI/tolk-flash", 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 RaspizdAI/tolk-flash with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/RaspizdAI/tolk-flash

SGLang

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

Docker Model Runner
How to use RaspizdAI/tolk-flash with Docker Model Runner:
```
docker model run hf.co/RaspizdAI/tolk-flash
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

tolk-flash

tolk-flash is a lightweight Flash version of the tolk-7B language model, based on Qwen 2.5 3B.

The model is designed for fast multi-turn conversational dialogue, persona alignment, and text generation in Russian and English while maintaining a small parameter count for efficient inference.

📌 Model Summary

Model Name: tolk-flash

Parameters: 3B

Base Model: unsloth/Qwen2.5-3B-bnb-4bit

Languages: Russian (ru), English (en)

Primary Task: Conversational Text Generation / Persona-based Dialogue

🚀 Quick Start

You can run tolk-flash using standard Hugging Face Transformers:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "RaspizdAI/tolk-flash"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

messages = [
    {"role": "user", "content": "Привет! Расскажи немного о себе."}
]

inputs = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt"
).to("cuda")

outputs = model.generate(
    input_ids=inputs,
    max_new_tokens=1024,
    temperature=0.2,
    top_p=0.9,
    do_sample=True
)

response = tokenizer.decode(
    outputs[0][inputs.shape[1]:],
    skip_special_tokens=True
)

print(response)

⚡ Why Flash?

tolk-flash is intended as a smaller and faster alternative to larger versions of the tolk-7B model.

With 3B parameters, it requires significantly fewer resources and is better suited for:

Fast inference
Weak GPUs
Low-resource environments
High-throughput conversational applications

📜 License

This model is made available under the MIT License. You are free to use, modify, distribute, and incorporate this model into commercial or non-commercial applications.