fix: wrong usage computation
4dad7a7
Morgan Funtowicz
commited on
Commit History
fix(import): remove unused imports
a6ada40
Morgan Funtowicz
commited on
feat(embedding): move to common I/O representation
49b79e1
Morgan Funtowicz
commited on
feat(http): move from openai only to http frontend package
5460614
Morgan Funtowicz
commited on
feat(embeddings): do not tokenize twice
5e1abf0
Morgan Funtowicz
commited on
feat(embeddings): flush denormal numbers
6ce5654
Morgan Funtowicz
commited on
misc(embeddings): use attention_mask to effectively compute the number of used tokens
276cf66
Morgan Funtowicz
commited on
misc(config): add proper way to detect if cpu may support bfloat16
a8540ed
Morgan Funtowicz
commited on
misc(embeddings): clean up
90c13c1
Morgan Funtowicz
commited on
feat(embeddings): fix invalid nested array when allocating the embedding responses
cb9be17
Morgan Funtowicz
commited on
feat(embeddings): expose some more to Python and return corresponding embedding (with copy for now)
38fa9fc
Morgan Funtowicz
commited on
feat(embeddings): allow the copy of vector elements (optimize later)
69894ec
Morgan Funtowicz
commited on
feat(embeddings): expose model name in the handler
e5ab730
Morgan Funtowicz
commited on
feat(embeddings): log out version of the SDK
1224145
Morgan Funtowicz
commited on