Upload folder using huggingface_hub

.gitattributes

@@ -58,3 +58,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+dataset_full.jsonl filter=lfs diff=lfs merge=lfs -text
+dataset_permissive.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,91 +1,31 @@
----
-language:
-- en
-- code
-license: other
-task_categories:
-- text2text-generation
-- text-generation
-tags:
-- opengl
-- glsl
-- webgl
-- shaders
-- computer-graphics
-- 3d
-- vulkan
-- code-generation
-size_categories:
-- 1K<n<10K
----
-
-# GLSL/OpenGL Educational Dataset for AI Training
-
-A curated, statically validated, and educational dataset of GLSL shaders, OpenGL 3.3/4.x programs, WebGL/WebGL2 demos, and procedural graphics pipelines.
-
-## Key Features
-
-- **Multi-Stage GLSL**: Vertex, Fragment, Compute, Geometry, and Tessellation shaders.
-- **Static Validation**: All shaders validated against Khronos `glslangValidator` (GLSL 1.00 ES up to GLSL 4.60 & SPIR-V).
-- **Inline Extraction**: Shaders extracted from C++ host code, raw string literals, `ssgl` lambdas, and WebGL JavaScript/HTML applications.
-- **Rich Annotations**: Includes GLSL dialect version, comment density, topic tagging (PBR, raymarching, lighting, compute, etc.), and complexity classification.
-- **License Partitioning**: Permissive (MIT, Apache-2.0, BSD, CC0) vs Non-Commercial partitions.
-- **Visual Previews**: High-quality screenshot previews linked to shader code.
-
-## Dataset Structure & Splits
-
-- `dataset_permissive.jsonl`: Clean, permissive open source licenses (MIT, Apache 2.0, BSD, CC0, Unlicense, Zlib).
-- `dataset_noncommercial.jsonl`: Non-commercial educational content (CC-BY-NC-4.0, Shadertoy CC-BY-NC-SA-3.0).
-- `dataset_full.jsonl`: Complete aggregated corpus.
-
-## Schema Specification
-
-Each record in JSONL contains:
-
-```json
-{
-  "id": "webgl_examples_bumpmapping",
-  "source": "https://github.com/damdoy/webgl_examples",
-  "source_commit": "b2c0c53...",
-  "collected_at": "2026-08-17T13:02:12+00:00",
-  "source_type": "repo",
-  "title": "Bumpmapping",
-  "api": "WebGL",
-  "glsl_version": "100",
-  "topic": "raymarching/texturing/bumpmapping/basics",
-  "difficulty": "advanced",
-  "files": [
-    {
-      "path": "bumpmapping/plane.js#[inline_1_vertex]",
-      "language": "glsl",
-      "stage": "vertex",
-      "loc": 18,
-      "comment_density": 0.0,
-      "code": "attribute vec4 vertex_pos;\n...",
-      "validation_status": "valid"
-    }
-  ],
-  "validation": {
-    "glslang_valid": 2,
-    "invalid": 0,
-    "unverified": 0,
-    "has_ubo_or_ssbo": false,
-    "texture_samplers": 1
-  },
-  "preview_image": "images/webgl_examples_bumpmapping.png",
-  "license": "MIT",
-  "non_commercial": false,
-  "comment_density": 0.054,
-  "dedup_hash": "4efd6250c8b26922"
-}
-```
-
-## Usage Example
-
-```python
-from datasets import load_dataset
-
-dataset = load_dataset("json", data_files={"train": "dataset_permissive.jsonl"})
-print(dataset["train"][0]["title"])
-print(dataset["train"][0]["files"][0]["code"])
-```
+---
+language:
+- en
+- code
+license: other
+task_categories:
+- text-generation
+tags:
+- opengl
+- glsl
+- webgl
+- shaders
+- computer-graphics
+- 3d
+- vulkan
+- code-generation
+size_categories:
+- 1K<n<10K
+---
+
+# GLSL/OpenGL Educational Dataset (AI Training Corpus)
+
+Aggregated across 2,096 curated samples from 42 major graphics repositories (OpenGL Core 3.3-4.6, WebGL/WebGL2, Vulkan GLSL/SPIR-V, Lygia shader modules, GL-Transitions, and procedural graphics pipelines).
+All shaders are statically verified with Khronos `glslangValidator`.
+
+## Dataset Statistics
+- **Total Samples**: 2,096
+- **Permissive Partition**: 1,965 (`dataset_permissive.jsonl`)
+- **Non-Commercial Partition**: 131 (`dataset_noncommercial.jsonl`)
+- **Total Shaders**: 10,364
+- **Valid GLSL Rate**: 7,802 / 10,364