Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,33 @@
+# Script Parser Gemma3 270M — ONNX
+
+Fine-tuned [google/gemma-3-270m-it](https://huggingface.co/google/gemma-3-270m-it) for structured script/screenplay parsing.
+
+Converts raw script text into pipe-delimited format:
+- `D|CHARACTER|DIRECTION|Line of dialogue`
+- `S|Scene heading text`
+- `C|CHARACTER|M/F/U` (gender classification)
+
+## Usage with transformers.js
+
+```javascript
+import { AutoTokenizer, AutoModelForCausalLM } from "@huggingface/transformers";
+
+const tokenizer = await AutoTokenizer.from_pretrained("frissonlabs/script-parser-gemma3-270m-ONNX");
+const model = await AutoModelForCausalLM.from_pretrained("frissonlabs/script-parser-gemma3-270m-ONNX", {
+  dtype: "q8",
+  device: "wasm", // or "webgpu"
+});
+```
+
+## Model Details
+
+- **Base model**: google/gemma-3-270m-it
+- **Fine-tuning**: QLoRA r=32, 10,262 training examples, 3 epochs
+- **Task**: Script/screenplay text → structured pipe-delimited format
+- **Eval**: 98% format accuracy on held-out eval set
+- **Export**: ONNX via optimum 2.1.0
+
+## Files
+
+- `onnx/model.onnx` — FP32 (1.6 GB)
+- `onnx/model_q8.onnx` — INT8 quantized (417 MB)