Upload model README

README.md

@@ -1,128 +1,81 @@
-## sm_llm Workflow
+# SM Coder 1.5B
 
-This folder contains the end-to-end pipeline to fine-tune `Qwen/Qwen2.5-Coder-1.5B-Instruct`
-on SuperMongo examples and export a quantized GGUF model for Ollama.
+SM Coder is a fine-tuned coding assistant focused on SuperMongo (SM) macro development for scientific plotting workflows.
 
-### Files
+This README is intended as the public model card for Hugging Face uploads.
 
-- `1_dataset.jsonl`: ChatML training set
-- `1b_alignment.jsonl`: targeted alignment examples (identity + anti-pseudo-syntax)
-- `2_finetune_unsloth.py`: LoRA fine-tuning (Unsloth)
-- `3_merge_weights.py`: merge LoRA into full model
-- `4_convert_gguf.py`: HF -> GGUF + Q4_K_M quantization
-- `5_build_mlc.py`: build MLC/WebLLM artifacts from merged model
-- `6_publish_hf.py`: publish GGUF and/or MLC artifacts to Hugging Face
-- `Modelfile.finetuned`: Ollama model definition
+## Model Repositories
 
-### Prerequisites (WSL + NVIDIA)
+- GGUF (Ollama / llama.cpp): https://huggingface.co/xpol555/sm-coder-gguf
+- MLC (WebLLM / browser deployment): https://huggingface.co/xpol555/sm-coder-mlc
 
-Run training from Linux/WSL with CUDA available.
+## What Is SuperMongo
 
-```bash
-nvidia-smi
-python3 -c "import torch; print(torch.cuda.is_available())"
-```
-
-### Environment Setup
-
-From `sm_llm/`:
-
-```bash
-uv sync --extra train
-```
-
-If you are using an already active venv, use `--active` in `uv run`.
-
-### Build llama.cpp (submodule)
-
-The conversion script expects `llama.cpp` at `./llama.cpp` (submodule path).
-
-```bash
-cd llama.cpp
-cmake -B build
-cmake --build build -j
-cd ..
-```
-
-### Train + Export
-
-```bash
-uv run --active --extra train python 2_finetune_unsloth.py
-uv run --active --extra train python 3_merge_weights.py
-uv run --active --extra train python 4_convert_gguf.py
-```
+SuperMongo is a plotting and data-analysis environment widely used in astronomy and scientific workflows.
 
-Generated artifacts:
+Reference:
+- https://www.astro.princeton.edu/~rhl/sm/
 
-- `lora_model/`
-- `merged_model/`
-- `gguf/sm-coder-1.5b-q4_k_m.gguf`
+## Dataset and Scope
 
-Note: `2_finetune_unsloth.py` loads both `1_dataset.jsonl` and `1b_alignment.jsonl` if present.
+This model was developed on a curated dataset of SM macros and related examples collected and evolved over about 20 years of real usage.
 
-### Ollama Test
+The training data emphasizes:
+- reusable plotting macros
+- parameterized workflow helpers
+- practical data visualization patterns used in long-running scientific analysis
 
-```bash
-ollama create sm-coder -f Modelfile.finetuned
-ollama run sm-coder
-```
-
-### Build MLC (WebLLM)
+## Intended Use
 
-Install MLC CLI (inside WSL, in your active env).
-The packages are hosted on a custom index and require platform-specific names.
-For CUDA 13.0 (RTX 4080 / sm89):
+- writing and refactoring SuperMongo macros
+- translating plotting requests into SM macro skeletons
+- helping document and modernize legacy macro collections
 
-```bash
-uv pip install --pre -U --find-links https://mlc.ai/wheels mlc-llm-nightly-cu130 mlc-ai-nightly-cu130
-```
+## Not Intended As
 
-> Note: `uv`-created venvs do not include `pip` by default, so use `uv pip` instead of `python -m pip`.
-> For other CUDA versions replace `cu130` with e.g. `cu128` or `cpu`.
+- a general-purpose chatbot
+- an authoritative source on non-SM domains
+- a replacement for validation/testing of scientific plots
 
-Build MLC artifacts from `merged_model/`:
+## Example Prompts
 
-```bash
-uv run --active python 5_build_mlc.py --model-id your-user/sm-coder-1.5b
-```
+- "Write an SM macro to draw concentric circles with configurable center and step."
+- "Create an SM macro that computes min/max from two vectors and sets plotting limits."
+- "Refactor this old SM macro into smaller parameterized helpers."
 
-Output is written to:
+## Example Macro Ideas
 
-- `mlc_dist/weights`
-- `mlc_dist/lib/*.wasm`
+The model can propose macro structures like these (adapt names and details to your local SM setup):
 
-### Publish on Hugging Face
+```sm
+define draw_circle (xc,yc,r,npt) {
+	set _th = 2*3.14159265*indgen(npt)/npt
+	set _x = xc + r*cos(_th)
+	set _y = yc + r*sin(_th)
+	connect _x _y
+}
 
-Install publish dependencies:
-
-```bash
-uv sync --extra publish
+define draw_concentric (xc,yc,r0,dr,nr,npt) {
+	do i = 0, nr-1 {
+		set _r = r0 + i*dr
+		draw_circle(xc,yc,_r,npt)
+	}
+}
 ```
 
-Login once:
-
-```bash
-huggingface-cli login
+```sm
+define autoscale_xy (xv,yv,pad) {
+	set _xmin = min(xv)
+	set _xmax = max(xv)
+	set _ymin = min(yv)
+	set _ymax = max(yv)
+	limits (_xmin-pad) (_xmax+pad) (_ymin-pad) (_ymax+pad)
+	box
+}
 ```
 
-Publish only GGUF (the model used by Ollama):
-
-```bash
-uv run --active --extra publish python 6_publish_hf.py \
-	--gguf-repo your-user/sm-coder-gguf
-```
-
-Publish both GGUF and MLC artifacts:
-
-```bash
-uv run --active --extra publish python 6_publish_hf.py \
-	--gguf-repo your-user/sm-coder-gguf \
-	--mlc-repo your-user/sm-coder-mlc
-```
-
-Use `--private` if you want private repositories.
-
-### Notes
+## Notes
 
-- On `/mnt/c` you may see UV hardlink warnings; they are harmless.
-- You can silence them with: `export UV_LINK_MODE=copy`
+- Validate generated macros against your SM version and local macro library.
+- For browser usage, prefer the MLC repository.
+- For local inference with Ollama/llama.cpp, prefer the GGUF repository.