Skip to content

Automation foundation model for tiny devices: 2-bit, 8-29 MB, tool calls, ASR, structured extraction and embeddings on phones, wearables, smart homes, robots, cars and microcontrollers.

5k validation only

HenryNdubuakuaad2d0967f34
diff --git a/src/data.py b/src/data.py
index aaf5b5e..3a33e72 100644
--- a/src/data.py
+++ b/src/data.py
@@ -392,22 +392,33 @@ def get_tokenizer(max_samples=None):
def load_tool_calls(split="train", max_samples=None, return_global_indices=False):
- """Load tool-calling dataset, splitting 90/10 for train/val.
+ """Load tool-calling dataset with shuffled train/val split.
+
+ Shuffles the full dataset with a fixed seed, then splits:
+ - val: min(5000, 10% of total)
+ - train: everything else
If return_global_indices is True, also return a numpy array mapping each
split-local row position back to its row id in the full unified dataset.
"""
ds = _load_unified_dataset()
n = len(ds)
+
+ rng = np.random.RandomState(42)
+ perm = rng.permutation(n)
+
+ val_size = min(5000, int(n * 0.1))
+ val_indices = perm[:val_size]
+ train_indices = perm[val_size:]
+
if split in ("validation", "val", "test"):
- start, end = int(n * 0.9), n
+ global_indices = val_indices.astype(np.int64)
elif split == "train":
- start, end = 0, int(n * 0.9)
+ global_indices = train_indices.astype(np.int64)
else:
- start, end = 0, n
+ global_indices = perm.astype(np.int64)
- global_indices = np.arange(start, end, dtype=np.int64)
- ds = ds.select(range(start, end))
+ ds = ds.select(global_indices.tolist())
if max_samples:
limit = min(max_samples, len(ds))