Views
No views yet
google/gemma-3-4b-it. The PLTs are TopK-48, expansion factor 64, and were
post-trained for 57 million tokens (run fingerprint 29b8c94c23cd846e).stats/layer_XX.safetensors: packed per-feature activation statistics.inputs/part-XXXXX.parquet: normalized input/token records referenced by the
activation tensors.manifest.json: shapes, provenance, file hashes, and join metadata.feature_ids, slot_counts, top_values, input_ids,
token_positions, and firing_counts. top_values are BF16; IDs, positions,
and slot counts are INT32; firing counts are INT64. Packed example arrays are
ordered by feature, with boundaries given by the cumulative sum of
slot_counts.input_ids join to the input_id column in the Parquet files. Those records
preserve the exact processed token IDs, token masks, optional multimodal token
types, formatted text, example identity, and source metadata.image_references contains stable pointers to the
source images, so consumers can resolve images using their own dataset access.1from huggingface_hub import hf_hub_download
2from safetensors import safe_open
3
4path = hf_hub_download(
5 "Jingcheng/gemma3-4b-it-plt-activations",
6 "stats/layer_12.safetensors",
7)
8with safe_open(path, framework="pt", device="cpu") as f:
9 feature_ids = f.get_tensor("feature_ids")
10 slot_counts = f.get_tensor("slot_counts")
11 top_values = f.get_tensor("top_values")
12 input_ids = f.get_tensor("input_ids")
13 token_positions = f.get_tensor("token_positions")
14 firing_counts = f.get_tensor("firing_counts")manifest.json for the exact model revision, source-checkpoint SHA-256,
per-file hashes, feature counts, and retained-example counts.