Phi-1.5 is a large language model released by Microsoft, released under the MIT license, with a claimed knowledge cutoff of 2023-09-11. Its weights, configuration, and tokenizer files have been cryptographically hashed and are queued for on-chain attestation by Weight Registry.
Description
Microsoft's 1.3B-parameter small language model trained almost entirely on curated 'textbook-quality' synthetic data (GPT-3.5-generated explanations and stories) rather than raw web-scale text, demonstrating that data quality can substitute for scale -- matches or beats models several times its size on common-sense and language-understanding benchmarks.
Metadata
Hugging Face stats
63,291 downloads on Hugging Face · 1,363 likes · pipeline: text-generation · library: transformers
Weight Hashes
Weights file
SHA-256b4f562cc4b12f1154a0b658b9d74ab9d024a686420fdf154ebc00a2219976df7
SHA3-2567d79d6141d3adb06173abd24c330f445766a07a9806ba58ce13ae16e404647e8
BLAKE30b805cd343d00fa3c675f36a24b65ffa3c745b5fb63891aeb80c97cb26fbaedc
Config file
SHA-2562fe52126434075e8a23aca5bfed8cca8aff7051779e6f11c76de9e9e9923dcef
SHA3-256a55a4056c66d2553b52f73921dd92bfd84ef3abcb5f31c49644abeb35cd3ef22
BLAKE34cd1e4707690248f45e4f39daf0e66dbc1c54c116dae26bfda34293876081ae1
Tokenizer file
SHA-256337da36be7a71a6e88aa9148967a7bc8736f4b47c7de8e19ba92b89e80734cfc
SHA3-2565afe567dc9fcdb1ad8c6f965c70c78288ee7312c6e0d27204e2b409b764df671
BLAKE3217d3dafd71c44b34cec26f21af6b1badfe7a4cd12b2a56ea43fdf38e1d1bd01
Verify it yourself
curl -L -o "model.safetensors" "https://huggingface.co/microsoft/phi-1_5/resolve/77aa61eeac94fbf33d492b9f2744c98b42d5b5eb/model.safetensors" sha256sum "model.safetensors"
Expected SHA-256: b4f562cc4b12f1154a0b658b9d74ab9d024a686420fdf154ebc00a2219976df7