Update README.md

README.md

@@ -1,3 +1,186 @@
----
-license: cc-by-4.0
----
+---
+license: cc-by-4.0
+---
+---
+language:
+  - en
+license: other
+license_name: License TBD — replace before publication
+pretty_name: H0 MatHamiltonian (~1 GiB WebDataset preview)
+tags:
+  - materials-science
+  - quantum-chemistry
+  - graph-neural-networks
+  - hamiltonian
+  - electronic-structure
+  - webdataset
+size_categories:
+  - n<1K
+dataset_info:
+  dataset_summary: |
+    ~1 GiB preview shard of a larger Hamiltonian dataset stored as WebDataset tar files.
+    Each sample is a pickled torch_geometric.data.Data blob (same encoding as the upstream LMDB release).
+---
+
+# Dataset Card for H0 MatHamiltonian — 1 GiB WebDataset Preview
+
+This card describes the **`hub_release_1gb_webdataset`** bundle: WebDataset-style **`train-*.tar`** shards plus **`train-manifest.csv`**, produced from the reviewer **~1 GiB LMDB** subset.
+
+## Dataset Details
+
+### Dataset Description
+
+This dataset is a **small, byte-limited preview** intended for **peer review, sanity checks, and pipeline testing**. It contains **45 crystal/material structures** as serialized graph/tensor records (Hamiltonian and overlap blocks, geometry, and metadata). The **full merged corpus** is much larger (~4.1 TB on disk in the upstream project); this Hub release is **not** the full dataset.
+
+- **Curated by:** *[Replace with author names and affiliations]*
+- **Funded by [optional]:** *[Replace or remove]*
+- **Shared by [optional]:** *[Replace or remove]*
+- **Language(s) (NLP):** Not applicable (structured scientific tensors / graphs; card text in English).
+- **License:** *[Replace with your chosen license, e.g. CC-BY-4.0, Apache-2.0, or custom]* — must match the legal terms you publish.
+
+### Dataset Sources [optional]
+
+- **Repository:** *[Replace with GitHub / project URL when public]*
+- **Paper [optional]:** *[Replace with citation when available]*
+- **Demo [optional]:** *[Optional notebook or Colab]*
+
+## Uses
+
+### Direct Use
+
+- Training or benchmarking **graph neural networks** and related models on **Hamiltonian / overlap tensor blocks** paired with **crystal structure** fields.
+- **Dataset loader development** and **integration tests** before downloading larger subsets.
+- **Reproducibility checks** during peer review (inspect manifests, decode samples).
+
+### Out-of-Scope Use
+
+- **Not** a guaranteed statistically representative sample of the full corpus (structure count is small; sampling follows upstream **index-list order** until a byte budget is reached).
+- **Not** for safety-critical or high-stakes decisions without domain validation.
+- **Not** a substitute for the full release when training production-scale models.
+
+## Dataset Structure
+
+### Files in this release
+
+| File / pattern | Description |
+|----------------|-------------|
+| `train-000000.tar`, `train-000001.tar`, … | WebDataset-style archives. Each logical sample has two members sharing the same basename. |
+| `00000000.pkl`, … | **Pickle blob** of `torch_geometric.data.Data` (same bytes as stored in the source LMDB value). |
+| `00000000.json`, … | Small JSON sidecar: `lmdb_key` (integer database index), `value_bytes` (byte length of `.pkl`). |
+| `train-manifest.csv` | Table mapping `sample_index` → shard file, member names, `lmdb_key`, `value_bytes`. |
+
+### Statistics (this build)
+
+- **Samples:** 45 (rows in `train-manifest.csv`).
+- **Shards:** 2 (`train-000000.tar`, `train-000001.tar`).
+- **Approximate on-disk tar footprint:** ~568 MiB + ~490 MiB (see files after upload; exact sizes depend on filesystem).
+
+### Example record fields
+
+After `pickle.loads` on a `.pkl` file, typical `Data` attributes include (names may vary slightly by pipeline version):
+
+- **Geometry / graph:** `pos`, `atoms`, `lattice`, `multi_edge_index`, `multi_edge_vec`, `lattice_translation_vector`, …
+- **Hamiltonian / overlap:** `diagonal_hamiltonian`, `off_diagonal_hamiltonian`, optional `diagonal_overlap`, `off_diagonal_overlap`, with associated `*_mask` tensors.
+- **Metadata:** `database_idx`, `mp_id`, `raw_basename`, `fermi_level`, `elapsed_time`, …
+
+Use `torch_geometric` introspection or print `data.keys` after loading.
+
+### Splits
+
+- **Single split:** all samples are shipped under the `train-*` shard prefix (preview bundle). There is **no** separate val/test split in this artifact.
+
+## Dataset Creation
+
+### Curation Rationale
+
+Provide a **compact, Hub-native format** (WebDataset `.tar`) for reviewers and integrators while preserving the **exact serialized sample representation** used in the LMDB-based pipeline.
+
+### Source Data
+
+#### Data Collection and Processing
+
+1. **Upstream:** Structures and electronic-structure-derived tensors originate from computational workflows (OpenMX-class outputs) merged into a large LMDB in the parent project.
+2. **Subset:** A **~1 GiB LMDB** subset was built by copying records in **canonical index order** until the byte budget was reached (same convention as documented in the upstream repo).
+3. **Conversion:** `convert_lmdb_to_webdataset.py` streamed LMDB cursor rows into `.tar` shards; each value became `{idx}.pkl`, with `{idx}.json` metadata.
+
+#### Who are the source data producers?
+
+Computational simulations and dataset merges performed by the **dataset authors / consortium** *[replace with specifics]*.
+
+### Annotations [optional]
+
+No separate human annotation layer beyond simulation-derived tensors and structured metadata.
+
+#### Personal and Sensitive Information
+
+This dataset is **not expected** to contain personal data. Identifiers such as `mp_id` refer to **materials database IDs**, not individuals.
+
+## Bias, Risks, and Limitations
+
+- **Coverage:** The preview contains **very few** structures relative to the full corpus; metrics are **not** indicative of full-dataset diversity.
+- **Simulation bias:** Electronic-structure workflows, basis choices, and convergence criteria induce systematic differences versus experiment.
+- **Technical:** `.pkl` relies on Python pickling; loading requires compatible **`torch` / `torch_geometric`** versions with the training stack used to build the data.
+
+### Recommendations
+
+- Validate loader behavior on this preview before scaling to multi-TB subsets.
+- Pin dependency versions in your training repo.
+- Prefer citing the **paper + dataset version + manifest checksum** once available.
+
+## Citation [optional]
+
+**BibTeX:**
+
+```bibtex
+% [Replace when the dataset paper / DOI is available]
+@misc{h0_mat_ham_2026,
+  title        = {H0 MatHamiltonian Dataset},
+  author       = {[Authors]},
+  year         = {2026},
+  howpublished = {\url{https://huggingface.co/datasets/[YOUR_ORG]/[YOUR_REPO]}}
+}
+```
+
+**APA:**
+
+*[Replace when citation details are finalized]*
+
+## Glossary [optional]
+
+- **LMDB:** Embedded key-value store used upstream; keys are 4-byte big-endian integers (`database_idx`).
+- **WebDataset / tar shards:** Convention of packing many small files into `.tar` archives for scalable streaming I/O.
+- **`torch_geometric.data.Data`:** PyG container holding tensor attributes and metadata for one graph/sample.
+
+## More Information [optional]
+
+- Conversion script (upstream repo): `convert_lmdb_to_webdataset.py`
+- Quick inspection script (upstream repo): `quick_check.py` (for LMDB; for this bundle, use tar + `pickle.loads` as below).
+
+### Minimal load example (single sample)
+
+```python
+import io
+import json
+import pickle
+import tarfile
+
+import torch  # noqa: F401 — needed for unpickling tensors inside Data
+
+tar_path = "train-000000.tar"
+with tarfile.open(tar_path, "r") as tar:
+    pkl_info = tar.getmember("00000000.pkl")
+    pkl_bytes = tar.extractfile(pkl_info).read()
+    meta = json.loads(tar.extractfile(tar.getmember("00000000.json")).read())
+
+data = pickle.loads(pkl_bytes)
+print(meta)   # {'lmdb_key': ..., 'value_bytes': ...}
+print(data)   # torch_geometric.data.Data
+```
+
+## Dataset Card Authors [optional]
+
+*[Replace — names of people who wrote this card]*
+
+## Dataset Card Contact
+
+*[Replace — email or issue tracker URL]*