Instructions to use Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 with Transformers:

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

pipe = pipeline("text-generation", model="Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5")
model = AutoModelForCausalLM.from_pretrained("Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5

SGLang

How to use Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 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 "Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5" \
    --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": "Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5",
		"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 "Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5" \
        --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": "Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Studio new

How to use Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 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 Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 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 Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5",
    max_seq_length=2048,
)

Docker Model Runner
How to use Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5 with Docker Model Runner:
```
docker model run hf.co/Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1.5
```

HyperThinkCode-Qwen3-8B-v1.5

File size: 1,822 Bytes

6ab4bc1
 
 
 
 
 
 
 
c6ca504
 
 
bc48c43
6ab4bc1
 
 
c6ca504
 
 
 
 
 
 
6ab4bc1
 
bc48c43
 
 
6ab4bc1
c6ca504
 
 
 
b924dd4
 
 
c6ca504
 
 
 
b924dd4
bc48c43
c6ca504
 
 
 
 
b924dd4
c6ca504
b924dd4
 
 
bc48c43
c6ca504
 
 
 
b924dd4
c6ca504
 
 
 
 
 
 
b924dd4
bc48c43
c6ca504
 
 
 
 
 
 
 
 
8a3cc37
bc48c43
c6ca504
 
 
 
 
 
6ab4bc1
c6ca504
bc48c43
c6ca504

---
base_model: unsloth/qwen3-8b-unsloth-bnb-4bit
tags:
- text-generation-inference
- transformers
- unsloth
- qwen3
- trl
- qlora
- reasoning
- code
- hyperthinkcode
license: apache-2.0
language:
- en
datasets:
- Sashvat/HyperThink-X-Nvidia-Opencode-Reasoning-200K
metrics:
- humaneval
- gsm8k
library_name: adapter
pipeline_tag: text-generation
---

# HyperThinkCode-Qwen3-8B-v1

HyperThinkCode-Qwen3-8B-v1 is a LoRA fine-tune of the Qwen3-8B base model.

---

## 🛠 Experimental Setup

- Base model: Qwen3-8B  
- Hardware: dual Tesla T4 (16GB VRAM each)  
- 4-bit QLoRA with rank = 16 and alpha = 16  
- All linear layers:  
  - Attention: q, k, v, o  
  - MLP: gate, up, down  
- Training time: ~1 hour 17 minutes  
- Total steps: 50  

---

## 🧠 Dataset & Objective

Training on a specific 30k subset of the  
**Sashvat/HyperThink-X-Nvidia-Opencode-Reasoning-200K** dataset.

- Uses chat template with assistant response in the *thinking* field  
- Objective: encourage *thinking over direct response*  
- Sequence length limited to 4096 tokens (for code complexity + VRAM constraints)  

---

## 📉 Training Logs

With only 50 steps, the loss shows expected variance given model + dataset complexity.

| Step | Training Loss |
|------|--------------|
| 10   | 0.8177       |
| 25   | 0.7358       |
| 50   | 0.6785       |

- Global batch size: 8 (1 device × 8 gradient steps)

---

## 📊 Evaluation (Ongoing)

Currently running benchmarks using the **lm-eval** library:

- HumanEval (Coding)
- GSM8K (Math)

Comparisons are being made against the base model.

---

## 🔁 Reproduction

```python
from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "Andy-ML-And-AI/HyperThinkCode-Qwen3-8B-v1",
    max_seq_length = 4096,
    load_in_4bit = True,
)