app.py

"""dcode Gradio Space - Text to Gcode inference."""

import re
import gradio as gr
import torch
from transformers import AutoModelForSeq2SeqLM, AutoModelForCausalLM, AutoTokenizer

# Available models
MODELS = {
    "flan-t5-base (best)": "twarner/dcode-flan-t5-base",
}

# Machine limits
BOUNDS = {"left": -420.5, "right": 420.5, "top": 594.5, "bottom": -594.5}

# Cache loaded models
_model_cache = {}


def get_model(model_name: str):
    """Load and cache model."""
    if model_name not in _model_cache:
        model_id = MODELS[model_name]
        device = "cuda" if torch.cuda.is_available() else "cpu"
        dtype = torch.float16 if device == "cuda" else torch.float32
        
        tokenizer = AutoTokenizer.from_pretrained(model_id)
        
        if "gpt2" in model_id or "codegen" in model_id:
            model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=dtype).to(device)
        else:
            model = AutoModelForSeq2SeqLM.from_pretrained(model_id, torch_dtype=dtype).to(device)
        
        model.eval()
        _model_cache[model_name] = (model, tokenizer, device)
    
    return _model_cache[model_name]


def validate_gcode(gcode: str) -> str:
    """Clamp coordinates to machine bounds."""
    lines = []
    for line in gcode.split("\n"):
        corrected = line
        
        x_match = re.search(r"X([-\d.]+)", line, re.IGNORECASE)
        if x_match:
            x = float(x_match.group(1))
            x = max(BOUNDS["left"], min(BOUNDS["right"], x))
            corrected = re.sub(r"X[-\d.]+", f"X{x:.2f}", corrected, flags=re.IGNORECASE)

        y_match = re.search(r"Y([-\d.]+)", line, re.IGNORECASE)
        if y_match:
            y = float(y_match.group(1))
            y = max(BOUNDS["bottom"], min(BOUNDS["top"], y))
            corrected = re.sub(r"Y[-\d.]+", f"Y{y:.2f}", corrected, flags=re.IGNORECASE)

        lines.append(corrected)

    return "\n".join(lines)


def generate(prompt: str, model_name: str, temperature: float, max_tokens: int) -> str:
    """Generate gcode from prompt."""
    if not prompt or not prompt.strip():
        return "Enter a prompt to generate gcode"

    try:
        model, tokenizer, device = get_model(model_name)
        model_id = MODELS[model_name]
        
        inputs = tokenizer(prompt, return_tensors="pt", max_length=128, truncation=True)
        inputs = {k: v.to(device) for k, v in inputs.items()}

        with torch.no_grad():
            outputs = model.generate(
                **inputs,
                max_new_tokens=max_tokens,
                do_sample=True,
                temperature=temperature,
                top_p=0.9,
                pad_token_id=tokenizer.eos_token_id,
            )

        # For causal models, skip the input tokens
        if "gpt2" in model_id or "codegen" in model_id:
            gcode = tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
        else:
            gcode = tokenizer.decode(outputs[0], skip_special_tokens=True)
        
        gcode = validate_gcode(gcode)
        line_count = len(gcode.split("\n"))
        return f"; dcode output - {line_count} lines\n; Model: {model_name}\n; Machine validated\n\n{gcode}"
    except Exception as e:
        return f"; Error: {e}"


demo = gr.Interface(
    fn=generate,
    inputs=[
        gr.Textbox(label="Prompt", placeholder="drawing of a cat...", lines=2),
        gr.Dropdown(choices=list(MODELS.keys()), value="flan-t5-base (best)", label="Model"),
        gr.Slider(0.1, 1.5, value=0.8, label="Temperature", info="Higher = more creative"),
        gr.Slider(256, 2048, value=1024, step=256, label="Max Tokens"),
    ],
    outputs=gr.Code(label="Gcode", language=None, lines=25),
    title="dcode",
    description="**Text → Polargraph Gcode** | Generate machine-compatible gcode from natural language. [GitHub](https://github.com/Twarner491/dcode) | [Model](https://huggingface.co/twarner/dcode-flan-t5-base) | [Dataset](https://huggingface.co/datasets/twarner/dcode-polargraph-gcode)",
    examples=[
        ["drawing of a cat", "flan-t5-base (best)", 0.8, 1024],
        ["abstract spiral pattern", "flan-t5-base (best)", 0.9, 1024],
        ["simple house with chimney", "flan-t5-base (best)", 0.7, 512],
        ["portrait of a woman", "flan-t5-base (best)", 0.8, 1024],
    ],
    theme=gr.themes.Soft(primary_hue="emerald"),
    article="""
## About

dcode finetunes text-to-text models to directly output polargraph-compatible gcode from natural language descriptions.

**Training**: Flan-T5-base trained on 175,952 art-caption-gcode triplets for 20 epochs on H100.

**Machine Bounds**: X: ±420.5mm, Y: ±594.5mm | Pen servo: 40° (down) / 90° (up)

**License**: MIT
""",
)

if __name__ == "__main__":
    demo.launch()