110 lines
3.4 KiB
Markdown
110 lines
3.4 KiB
Markdown
---
|
||
library_name: transformers
|
||
tags:
|
||
- Physics
|
||
- Quantum
|
||
- reasoning
|
||
- unsloth
|
||
- sft
|
||
- lora
|
||
license: apache-2.0
|
||
datasets:
|
||
- mattwesney/CoT_Reasoning_Quantom_Physics_And_Computing
|
||
language:
|
||
- en
|
||
base_model:
|
||
- unsloth/Qwen3-1.7B
|
||
pipeline_tag: text-generation
|
||
---
|
||
|
||
# Quantum-ToT
|
||
|
||
## Model Details
|
||
|
||
Quantum-ToT is a fine-tuned variant of Qwen3-1.7B, optimized for Chain-of-Thought (CoT) reasoning in quantum mechanics and quantum computing contexts.
|
||
This model was trained using the [moremilk/CoT_Reasoning_Quantum_Physics_And_Computing](https://huggingface.co/datasets/moremilk/CoT_Reasoning_Quantom_Physics_And_Computing) dataset — a curated collection of question–answer pairs that go beyond surface-level definitions to show the logical reasoning process behind quantum concepts.
|
||
|
||
The goal of this fine-tuning is to enhance the model’s ability to:
|
||
|
||
- Explain quantum principles with structured, step-by-step logic
|
||
- Reason through conceptual problems in quantum physics and computing
|
||
- Support educational and research applications that require interpretable reasoning chains
|
||
|
||
|
||
## Uses
|
||
|
||
### Direct Use
|
||
|
||
- Educational assistance in quantum physics and quantum computing
|
||
- AI tutors or reasoning assistants for STEM learning
|
||
- Conceptual reasoning benchmarks involving quantum phenomena
|
||
- Research in reasoning-aware model behavior and CoT interpretability
|
||
|
||
### Out of Scope
|
||
|
||
- Predicting new or unverified physical phenomena
|
||
- Running quantum simulations or algorithmic derivations
|
||
- Hardware-level quantum design
|
||
- Real-time physics predictions
|
||
|
||
## Bias, Risks, and Limitations
|
||
|
||
- May hallucinate if prompted outside the quantum domain
|
||
- Not suitable for advanced quantum algorithm design or experimental predictions
|
||
|
||
## How to Get Started with the Model
|
||
|
||
Use the code below to get started with the model.
|
||
|
||
```python
|
||
from transformers import AutoTokenizer, AutoModelForCausalLM
|
||
|
||
tokenizer = AutoTokenizer.from_pretrained("khazarai/Quantum-ToT")
|
||
model = AutoModelForCausalLM.from_pretrained(
|
||
"khazarai/Quantum-ToT",
|
||
device_map={"": 0}
|
||
)
|
||
|
||
question = """
|
||
Explain the Heisenberg Uncertainty Principle in detail, including its mathematical formulation, physical implications, and common misconceptions.
|
||
"""
|
||
|
||
messages = [
|
||
{"role" : "user", "content" : question}
|
||
]
|
||
text = tokenizer.apply_chat_template(
|
||
messages,
|
||
tokenize = False,
|
||
add_generation_prompt = True,
|
||
enable_thinking = True,
|
||
)
|
||
|
||
from transformers import TextStreamer
|
||
_ = model.generate(
|
||
**tokenizer(text, return_tensors = "pt").to("cuda"),
|
||
max_new_tokens = 3000,
|
||
temperature = 0.6,
|
||
top_p = 0.95,
|
||
top_k = 20,
|
||
streamer = TextStreamer(tokenizer, skip_prompt = True),
|
||
)
|
||
```
|
||
|
||
### Dataset
|
||
|
||
Dataset: [moremilk/CoT_Reasoning_Quantum_Physics_And_Computing](https://huggingface.co/datasets/moremilk/CoT_Reasoning_Quantom_Physics_And_Computing)
|
||
|
||
This dataset contains rich reasoning-based question–answer pairs covering:
|
||
|
||
- Core quantum principles: superposition, entanglement, measurement
|
||
- Effects of quantum gates (Hadamard, Pauli-X/Y/Z, etc.) on qubits
|
||
- Multi-qubit reasoning (e.g., Bell states, entangled systems)
|
||
- Basic quantum algorithms and logical operations
|
||
- Probabilistic interpretation of measurement outcomes
|
||
|
||
Each entry includes:
|
||
|
||
- think block → model’s internal reasoning process
|
||
- answer block → final concise explanation or solution
|
||
|
||
The dataset focuses on conceptual understanding rather than heavy mathematical derivations or complex quantum hardware design. |