Upload README.md with huggingface_hub

README.md

@@ -1,35 +1,110 @@
 ---
-dataset_info:
-  features:
-  - name: problem_id
-    dtype: int64
-  - name: problem_title
-    dtype: string
-  - name: difficulty
-    dtype: string
-  - name: c_source
-    dtype: string
-  - name: source_repo
-    dtype: string
-  - name: architecture
-    dtype: string
-  - name: optimization
-    dtype: string
-  - name: compiler
-    dtype: string
-  - name: assembly
-    dtype: string
-  - name: compilation_success
-    dtype: bool
-  splits:
-  - name: train
-    num_bytes: 49696480
-    num_examples: 7056
-  download_size: 8106173
-  dataset_size: 49696480
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
+license: apache-2.0
+task_categories:
+  - text-generation
+language:
+  - en
+tags:
+  - assembly
+  - c
+  - leetcode
+  - compiler
+  - code
+size_categories:
+  - 1K<n<10K
 ---
+
+# LeetCode Assembly Dataset
+
+441 LeetCode problems solved in C, compiled to assembly across **4 architectures** and **4 optimization levels** using GCC via the [Godbolt Compiler Explorer](https://godbolt.org) API.
+
+## Dataset Summary
+
+| Stat | Value |
+|------|-------|
+| Total rows | 7,056 |
+| Unique problems | 441 |
+| Architectures | x86-64, AArch64, MIPS64, RISC-V 64 |
+| Optimization levels | -O0, -O1, -O2, -O3 |
+| Compiler | GCC 15.2 |
+| Compilation success rate | 100% |
+| Difficulty split | Easy: 98, Medium: 259, Hard: 84 |
+
+Each problem has **16 assembly variants** (4 architectures x 4 optimization levels), making this useful for studying how the same algorithm compiles differently across ISAs and optimization settings.
+
+## Schema
+
+| Column | Type | Description |
+|--------|------|-------------|
+| `problem_id` | int | LeetCode problem number |
+| `problem_title` | string | Problem name (e.g. "Two Sum") |
+| `difficulty` | string | Easy, Medium, or Hard |
+| `c_source` | string | Complete C source code |
+| `source_repo` | string | GitHub repo the solution came from |
+| `architecture` | string | Target ISA: `x86-64`, `aarch64`, `mips64`, `riscv64` |
+| `optimization` | string | GCC optimization flag: `-O0`, `-O1`, `-O2`, `-O3` |
+| `compiler` | string | Compiler version used |
+| `assembly` | string | Compiled assembly output |
+| `compilation_success` | bool | Whether compilation succeeded |
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("ronantakizawa/leetcode-assembly", split="train")
+
+# Get all x86-64 assembly at -O2
+x86_O2 = ds.filter(lambda r: r["architecture"] == "x86-64" and r["optimization"] == "-O2")
+
+# Compare the same problem across architectures
+two_sum = ds.filter(lambda r: r["problem_id"] == 1 and r["optimization"] == "-O2")
+for row in two_sum:
+    print(f"--- {row['architecture']} ---")
+    print(row["assembly"][:200])
+    print()
+```
+
+## Sources
+
+C solutions were collected from three open-source GitHub repositories of pure-C LeetCode solutions:
+
+- [vli02/leetcode](https://github.com/vli02/leetcode) (383 solutions)
+- [begeekmyfriend/leetcode](https://github.com/begeekmyfriend/leetcode) (215 solutions)
+- [lennylxx/leetcode](https://github.com/lennylxx/leetcode) (159 solutions)
+
+After deduplication, 441 unique problems remained. Solutions are self-contained C files using only the standard library (no C++ STL), producing clean and readable assembly output.
+
+## Compilation Details
+
+Assembly was generated using the [Godbolt Compiler Explorer](https://godbolt.org) public API with the following settings:
+
+| Architecture | Compiler ID | Syntax |
+|-------------|-------------|--------|
+| x86-64 | `cg152` (GCC 15.2) | Intel |
+| AArch64 | `carm64g1520` (GCC 15.2) | Native |
+| MIPS64 | `cmips64g1520` (GCC 15.2) | Native |
+| RISC-V 64 | `rv64-cgcc1520` (GCC 15.2) | Native |
+
+Compiler flags: `-std=gnu89 -w -include stdio.h -include stdlib.h -include string.h -include stdbool.h -include limits.h -include math.h -include stdint.h -include ctype.h`
+
+Assembly output filters: directives removed, comments removed, labels preserved, symbols demangled.
+
+## Potential Use Cases
+
+- Training or evaluating models on C-to-assembly translation
+- Studying how optimization levels affect generated code
+- Cross-architecture assembly comparison and analysis
+- Assembly language education and reference
+- Compiler behavior research
+
+## Build Pipeline
+
+The dataset was built with an automated Python pipeline:
+
+1. **Extract**: Clone repos, extract `.c` files, normalize unicode, deduplicate by problem ID
+2. **Metadata**: Enrich with LeetCode problem titles and difficulty from HuggingFace
+3. **Compile**: Send each solution to Godbolt API across all architecture/optimization combinations (with SQLite checkpointing for resumability)
+4. **Publish**: Build HuggingFace Dataset and push to Hub
+
+Source code for the pipeline is available at [github.com/ronantakizawa/leetcodeassembly](https://github.com/ronantakizawa/leetcodeassembly).