alexwengg's picture
Core ML fp16 L64/L128/L256/L512 + tokenizer + Evoke P2.2 pooling config
4ef29e0 verified
Raw History Blame Contribute Delete
350 Bytes
Core ML conversion of ibm-granite/granite-embedding-30m-sparse
(revision ad82b1fd09541c998c8d45045d601c51fdb8a9b7), Apache License 2.0, by IBM.
The query/document pooling constants in config.json come from Intelligent Internet's
Evoke model checkout Intelligent-Internet/Evoke-Model-Beta-1 (P2.2), Apache License 2.0.
Converted by Fluid Inference.