WestQuant Open Source

WQT50M — a 50M-parameter open-source Transformer for quantum representation scheduling. Trained on real Qiskit transpilation outputs.

AI schedules. Deterministic mathematics executes. Independent verification certifies.

WQT50M is a proof-of-concept that a 50M-parameter Transformer can learn structured quantum optimization preferences from real compiler outputs. Researchers building better compiler-optimization models can use this as a baseline.

Models

Model	Params	Training Data	Key Result
WQT20M-Beta	19M	Synthetic	Proof-of-concept baseline
WQT50M	50M	Real Qiskit	74% Top-1, 90.4% beats random

What WQT50M Does — 3 Tested Examples

Example 1: 62% Gate Reduction

Problem:  8-qubit random circuit (depth 30) on grid topology
          Naive transpilation → cost 1231.2 (fidelity-focused)

          WQT50M schedules ZX_SIMPLIFY
          → cost 467.4 (62.0% reduction)

          Random selection → cost 869.6 (29.4% reduction)
          Oracle (exhaustive) → cost 467.4 (62.0%)
          WQT50M matches oracle ✓

Example 2: Objective-Aware Scheduling

QAOA-6q-p2 on linear backend:

  Objective            WQT50M picks
  ─────────────────    ─────────────────
  balanced             → CANCEL_GATES
  depth_focused        → NATIVE_GATESET
  fidelity_focused     → FUSE_ROTATIONS
  time_focused         → MERGE_ADJACENT

  4 different actions for 5 objectives ✓

Example 3: Calibrated Value Prediction

Action:   ZX_SIMPLIFY
          Predicted cost: 467.4
          Actual cost:    467.4
          Error:          0.0%

Calibration ratio: 0.999 (predicted range matches actual range)
Spearman correlation: 0.981

---

Quick Start

pip install transformers torch

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained("WestQuantStudio/WQT50M")
tokenizer = AutoTokenizer.from_pretrained("WestQuantStudio/WQT50M")
device = "mps" if torch.backends.mps.is_available() else "cpu"
model = model.to(device).eval()

state = ("<DOMAIN:circuit_optimization> <LEVEL:CIRCUIT> "
         "<N_QUBITS:8> <ENTANGLEMENT:0.5000> "
         "<RES:n_q=8 D=30 G1=100 G2=20 T=120 M=0 A=0.2000 E=0.0050 C=0.5000> "
         "<BACKEND:SUPERCONDUCTING> <TOPO:grid> "
         "<T1:180us> <T2:90us> <READOUT_ERR:0.0120> "
         "<OBJ_TYPE:fidelity_focused> <STEP:0/5>")

text = f"<PREDICT> {state} <ACTION> ZX_SIMPLIFY <COST> "
ids = tokenizer.encode(text, add_special_tokens=False, return_tensors="pt").to(device)
with torch.no_grad():
    for _ in range(8):
        out = model(input_ids=ids)
        nxt = out.logits[0, -1].argmax().unsqueeze(0).unsqueeze(0)
        ids = torch.cat([ids, nxt], dim=1)
        if nxt.item() == tokenizer.eos_token_id:
            break

print(tokenizer.decode(ids[0].tolist()).split("<COST>")[-1].strip())

---

Validation Results

450 tests: 15 circuits × 5 objectives × 6 backends

Metric	WQT20M-Beta	WQT50M
Top-1 accuracy	~22%	74.0%
Beats random	58%	90.4%
Avg improvement over naive	-14.8%	+16.6%
Efficiency (% of oracle)	23.3%	88.7%
Value calibration ratio	~0.01	0.999
Objective sensitivity	0%	65%
Backend sensitivity	~0%	60%

---

Production Gap Coverage

Gap	WQT20M-Beta	WQT50M	Status
Real compiler outputs	Synthetic	Real Qiskit	✅
Value calibration	6.0–6.6	Ratio 0.999	✅
Objective sensitivity	0%	65%	✅
Legality	76.9%	79.2%	⚠️
Multi-step trajectories	Single-step	2-5 steps	✅
Backend awareness	Tokens only	60%	✅
Model capacity	19M	49.53M	✅
Search improvement	+35%	90.4% win	✅

---

Model

Config	Value
Architecture	Llama-style decoder Transformer
Parameters	49.53M
Layers	12
Hidden size	512
Attention heads	8 (4 KV heads, GQA)
FFN	2048 (SwiGLU)
Context length	4096
Vocab	4561 (quantum-native structured tokens + BPE)
Model size	189 MB

Models:

• WQT50M (50M, real Qiskit data)
• WQT20M-Beta (19M, synthetic baseline)

License

Apache 2.0