Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

ELF-B 1B samples (T5 ids)

Unconditional samples from the official ELF-B (github.com/lillian039/ELF, PyTorch branch), trained for 1B tokens on 9 shards of embedded-language-flows/openwebtext-t5. Raw (non-EMA) weights; self-conditioning CFG 3; seed 20260923; 128 samples each.

file sampler sha256
elfb1b_raw_sde32.npy SDE, 32 steps, gamma 1.5 d20313f65b0fec6f1cb2b8601fe41fccd3b9108427fd6e2fa55b4ae8c909e344
elfb1b_raw_sde64.npy SDE, 64 steps, gamma 1.0 88d52a201c67b38ceb1483607603b5b7f0d540acfae72fb70d01b46d03e2aafd

Each file is int32 [128, 1024] google-t5/t5-small token ids, untrimmed (EOS = 1). They are used for a length-matched GPT-2 Large comparison against a causal flow model.

Downloads last month
37