GPT-Neo-125M is a large language model released by EleutherAI, released under the MIT license, with a claimed knowledge cutoff of 2021-03-21. Its weights, configuration, and tokenizer files have been cryptographically hashed and are queued for on-chain attestation by Weight Registry.
Description
EleutherAI's open reproduction of GPT-3-class autoregressive transformers (125M, smallest of the GPT-Neo family which scales to 2.7B), trained on the 825GB Pile dataset and released while GPT-3 itself remained closed/API-only -- an early, direct open answer to OpenAI's non-release.
Metadata
Hugging Face stats
531,821 downloads on Hugging Face · 229 likes · pipeline: text-generation · library: transformers
Weight Hashes
Weights file
SHA-25652738cbfb54e25a232598242f60ef19ee193d36090b98fe649b10c02724b3521
SHA3-256369b0cc5eb4dbed108d1ad8ba588e4a32c8fd0015593c8672518db32a7c11f86
BLAKE3d85d9e4e87e3f5e7fa8a430e1a2330bb78d41f140354871bc5f218addc29e223
Config file
SHA-256dace197ce91788731063527367b0a6766d7d3a4ab72c671fabf16dbbf9037e16
SHA3-2567a9f0344f8b57c9e87879afa8a8187bb03a6cc156b18a68e56549018872e3444
BLAKE315d98a9f90746adef31ec7236394c33c9637b2ae9630fe8bb2fe0dbd73a96f85
Tokenizer file
SHA-256f6ed3d307010c244c22aeffbde05f419cf277c23e64cf98b673cac5449cfeff5
SHA3-256ff21232b1752dceb0049a0d95c6d7ac851636e987464509d1b87f31522b5ac2f
BLAKE3897261cf797a9f2e242b0ce7567a07db028328af71183642068dcdff38fb0983
Verify it yourself
curl -L -o "model.safetensors" "https://huggingface.co/EleutherAI/gpt-neo-125m/resolve/21def0189f5705e2521767faed922f1f15e7d7db/model.safetensors" sha256sum "model.safetensors"
Expected SHA-256: 52738cbfb54e25a232598242f60ef19ee193d36090b98fe649b10c02724b3521