-
Upload gpt_neox.layers.10.mlp.dense_4h_to_h_float32_mom2_100000.npz
-
Finetunes llama-2b resid-pre-16 SAEs on up to 100M tokens of Harry Potter
-
Trains llama-2b resid-pre-16 SAEs on up to 200M tokens of OpenWebText
-
Upload npz needed for rome edits
-
Initial commit
-
Fine-tunes resid_5_pre SAE on Harry Potter (70M tokens of fine-tuning)
-
Adds 300M resid_5_pre SAE trained on the pile
-
Initial commit
-
Initial commit
-
Initial commit
-
Initial commit
-
Initial commit
-
Upload folder using huggingface_hub
-
Upload folder using huggingface_hub
-
Upload folder using huggingface_hub
-
Upload folder using huggingface_hub
-
Upload folder using huggingface_hub
-
Upload model checkpoint
-
Upload model checkpoint
-
initial commit