Instructions to use dungnvt/Tini1.5-8B-A1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dungnvt/Tini1.5-8B-A1B with Transformers:

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

pipe = pipeline("text-generation", model="dungnvt/Tini1.5-8B-A1B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("dungnvt/Tini1.5-8B-A1B")
model = AutoModelForMultimodalLM.from_pretrained("dungnvt/Tini1.5-8B-A1B")
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 dungnvt/Tini1.5-8B-A1B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dungnvt/Tini1.5-8B-A1B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dungnvt/Tini1.5-8B-A1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dungnvt/Tini1.5-8B-A1B

SGLang

How to use dungnvt/Tini1.5-8B-A1B 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 "dungnvt/Tini1.5-8B-A1B" \
    --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": "dungnvt/Tini1.5-8B-A1B",
		"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 "dungnvt/Tini1.5-8B-A1B" \
        --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": "dungnvt/Tini1.5-8B-A1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio

How to use dungnvt/Tini1.5-8B-A1B with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dungnvt/Tini1.5-8B-A1B to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dungnvt/Tini1.5-8B-A1B to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for dungnvt/Tini1.5-8B-A1B to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="dungnvt/Tini1.5-8B-A1B",
    max_seq_length=2048,
)

Docker Model Runner
How to use dungnvt/Tini1.5-8B-A1B with Docker Model Runner:
```
docker model run hf.co/dungnvt/Tini1.5-8B-A1B
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Tini1.5-8B-A1B

Tini1.5-8B-A1B is a production-grade fine-tuned version of the hybrid model architecture LiquidAI/LFM2.5-8B-A1B. This model is highly optimized for English Agentic Reasoning, seamlessly combining deep multi-turn Chain-of-Thought (CoT) and strict policy-compliant native system function calling capabilities.

📊 Dataset Mixture

The model was Supervised Fine-Tuned (SFT) on a symmetric data matrix balancing deep reasoning tracks and advanced tool execution trajectories:

Dataset	Category
`nvidia/Nemotron-SFT-Agentic-v2`	Tool Use / Multi-step Agentic Policy
`Jackrong/DeepSeek-V4-Distill-8000x`	Pure Reasoning / Math / Code
`nohurry/Opus-4.6-Reasoning-3000x-filtered`	Advanced CoT Reasoning
`Jackrong/Qwen3.5-reasoning-700x`	Hard Logic & Complex Math

🛠️ Training Techniques

To preserve the model's core architecture while focusing gradient updates entirely on structured agent workflows, the following configurations were applied: Train on Response Only

🏃‍♂️ Quick Start & Inference Parameters Guide

💡 Recommended Decoding Parameters

Agentic & Function-Calling Tasks: temperature: 0.1 | top_p: 0.95 | top_k: 50 | repetition_penalty: 1.00
General Coding & Contextual Reasoning Tasks: temperature: 0.35 | top_p: 0.90 | top_k: 40 | repetition_penalty: 1.05

🚀 Python Example Script

import torch
from unsloth import FastLanguageModel
from transformers import TextStreamer

MODEL_PATH = "iselabvn/Tini1.5-8B-A1B"

# 1. Load model with 4-bit quantization and essential regex patches
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = MODEL_PATH,
    max_seq_length = 8192,
    dtype = torch.bfloat16,
    load_in_4bit = True,
    trust_remote_code = True,
    fix_mistral_regex = True
)
FastLanguageModel.for_inference(model)

# 2. Configure system prompt aligned with strict v1.5 validation guardrails
messages = [
    {
        "role": "system", 
        "content": "You are an advanced, high-efficiency executive Agent. If a tool requires a parameter that is missing from the prompt, DO NOT analyze or debate the schema. Stop thinking immediately and output a clear question asking the user for that parameter."
    },
    {
        "role": "user", 
        "content": "What is the current stock price of Nvidia (NVDA) today?"
    }
]

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

# 3. Generate structured output within a safe tokens boundary
with torch.no_grad():
    _ = model.generate(
        input_ids = inputs,
        streamer = text_streamer,
        max_new_tokens = 2048,
        use_cache = True,
        temperature = 0.1,
        top_p = 0.95,
        top_k = 50,
        repetition_penalty = 1.00
    )

Downloads last month: -

Safetensors

Model size

8B params

Tensor type

F32

BF16

Model tree for dungnvt/Tini1.5-8B-A1B

Base model

LiquidAI/LFM2.5-8B-A1B-Base

Finetuned

LiquidAI/LFM2.5-8B-A1B

Finetuned

(24)

this model

Quantizations

2 models

Duplicated from iselabvn/Tini1.5-8B-A1B

dungnvt
/

Tini1.5-8B-A1B