README.md

---
language:
  - en
license: apache-2.0
base_model: Qwen/Qwen2.5-Coder-1.5B-Instruct
base_model_relation: quantized
tags:
  - supermongo
  - astronomy
  - code-generation
  - scientific-plotting
  - webllm
pipeline_tag: text-generation
library_name: mlc-llm
---

# SM Coder 1.5B — MLC (WebLLM)

A fine-tuned coding assistant for **SuperMongo (SM)** macro development,
quantized to **q4f16_1** for browser inference via
[WebLLM](https://github.com/mlc-ai/web-llm).

Use it for free [here](https://huggingface.co/spaces/xpol555/supermongo-ai-webllm)!

## Usage with WebLLM

This repo contains only the quantized **weights** and `mlc-chat-config.json`.
Because the model uses the standard **Qwen2** architecture with **q4f16_1**
quantization, the pre-built wasm library shipped by `@mlc-ai/web-llm` works
out of the box — no custom compile step is needed.

```ts
import { CreateMLCEngine, prebuiltAppConfig } from "@mlc-ai/web-llm";

const engine = await CreateMLCEngine("xpol555/sm-coder-mlc", {
  appConfig: {
    ...prebuiltAppConfig,
    model_list: [
      ...prebuiltAppConfig.model_list,
      {
        model: "https://huggingface.co/xpol555/sm-coder-mlc/resolve/main/weights/",
        model_id: "xpol555/sm-coder-mlc",
        model_lib:
          prebuiltAppConfig.model_list.find(
            (m) => m.model_id === "Qwen2.5-1.5B-Instruct-q4f16_1-MLC"
          )!.model_lib,
      },
    ],
  },
});

const reply = await engine.chat.completions.create({
  messages: [{ role: "user", content: "Write an SM macro to draw error bars" }],
});
console.log(reply.choices[0].message.content);
```

## Base Model

[Qwen/Qwen2.5-Coder-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B-Instruct),
fine-tuned on a curated dataset of SM macros and related examples collected
over ~20 years of real scientific usage.

## What Is SuperMongo

SuperMongo is a plotting and data-analysis environment widely used in
astronomy and scientific workflows.

Reference: <https://www.astro.princeton.edu/~rhl/sm/>

## Intended Use

- Writing and refactoring SuperMongo macros.
- Translating plotting requests into SM macro skeletons.
- Helping document and modernize legacy macro collections.

## Limitations

- Not a general-purpose chatbot.
- Not an authoritative source on non-SM domains.
- Always validate generated macros against your SM version and local macro
  library before use.

## Example Output

The model produces macros in standard SM syntax (positional `$1 $2 …`
parameters, `set` for vectors, `define` for scalars):

```
draw_circle 4   ## draw a circle centred at ($1,$2) radius $3 with $4 points
        set _th = 2*PI*indgen($4)/$4
        set _cx = $1 + $3*COS(_th)
        set _cy = $2 + $3*SIN(_th)
        connect _cx _cy
```

```
quickplot 1     ## read two-column file $1, auto-scale, and plot
        data $1
        read {x 1 y 2}
        limits x y
        erase
        box
        connect x y
        xlabel Column 1
        ylabel Column 2
```

## Other Formats

- **GGUF (Ollama / llama.cpp):** <https://huggingface.co/xpol555/sm-coder-gguf>