Update README.md

README.md

@@ -25,7 +25,7 @@ tags:
 - crispr
 - cas9
 - open-data
-- nstruction-tuning
+- instruction-tuning
 pretty_name: STXBP1 ClinVar Curated Variants
 size_categories:
 - 10M<n<100M
@@ -151,4 +151,93 @@ _Main split for Hugging Face: JSONL format (see above for statistics)._
   "onc_fields": {},
   "sci_fields": {},
   "incl_fields": {}
-}
+}
+
+
+
+===================================================================================================================
+## You can easily load this dataset using the 🤗 Datasets library.
+
+The Hugging Face infrastructure will automatically use the efficient Parquet files by default, but you can also specify the JSONL if you prefer.
+
+### Install dependencies (if needed):
+
+```bash
+pip install datasets```
+
+
+
+## Load the full dataset (Parquet, recommended)
+
+```from datasets import load_dataset
+
+# This will automatically use the Parquet shards
+ds = load_dataset("SkyWhal3/ClinVar-STXBP1-NLP-Dataset")
+
+# Access examples
+print(ds["train"][0])```
+
+
+
+## To force JSONL loading (if you prefer the original format):
+
+```from datasets import load_dataset
+
+# Specify data_files to point to JSONL file(s)
+ds = load_dataset(
+    "SkyWhal3/ClinVar-STXBP1-NLP-Dataset",
+    data_files="ClinVar-STXBP1-NLP-Dataset.jsonl",
+    split="train"
+)
+print(ds[0])```
+
+
+## Other ways to use the data
+   Load all Parquet shards with pandas
+
+```import pandas as pd
+import glob
+
+# Load all Parquet shards in the train directory
+parquet_files = glob.glob("default/train/*.parquet")
+df = pd.concat([pd.read_parquet(pq) for pq in parquet_files], ignore_index=True)
+print(df.shape)
+print(df.head())```
+
+
+## Filter for a gene (e.g., STXBP1)
+
+```df = pd.read_parquet("default/train/0000.parquet")
+stxbp1_df = df[df["gene"] == "STXBP1"]
+print(stxbp1_df.head())```
+
+
+## Randomly sample a subset
+
+```sample = df.sample(n=5, random_state=42)
+print(sample)```
+
+
+## Load with Polars (for high performance)
+
+```import polars as pl
+
+df = pl.read_parquet("default/train/0000.parquet")
+print(df.head())```
+
+
+## Query with DuckDB (SQL-style)
+
+```import duckdb
+
+con = duckdb.connect()
+df = con.execute("SELECT * FROM 'default/train/0000.parquet' WHERE gene='STXBP1' LIMIT 5").df()
+print(df)```
+
+
+```## Streaming mode with 🤗 Datasets
+
+
+ds = load_dataset("SkyWhal3/ClinVar-STXBP1-NLP-Dataset", split="train", streaming=True)
+for record in ds.take(5):
+    print(record)```