Closed Wauplin closed 2 months ago
Related to https://github.com/keras-team/keras/pull/19854, https://github.com/huggingface/huggingface.js/pull/774, https://github.com/huggingface/hub-docs/pull/1320 and, https://github.com/huggingface/moon-landing/pull/10501 (private).
See https://github.com/huggingface/api-inference/pull/1956 for complete drop in private repo.
As discussed in https://github.com/huggingface/moon-landing/pull/10501#issuecomment-2196278636 (private), it's better to drop Keras models in Inference API, at least for now. This choice was motivated by the fact that:
keras
tf-keras
Given its low usage and scope, let's just drop the support. If keras becomes a major framework on the platform (which we hope :crossed_fingers: ), then we can reassess.
Related to https://github.com/keras-team/keras/pull/19854, https://github.com/huggingface/huggingface.js/pull/774, https://github.com/huggingface/hub-docs/pull/1320 and, https://github.com/huggingface/moon-landing/pull/10501 (private).
See https://github.com/huggingface/api-inference/pull/1956 for complete drop in private repo.
As discussed in https://github.com/huggingface/moon-landing/pull/10501#issuecomment-2196278636 (private), it's better to drop Keras models in Inference API, at least for now. This choice was motivated by the fact that:
keras
models will be tagged astf-keras
, thus breaking the current Inference APIGiven its low usage and scope, let's just drop the support. If
keras
becomes a major framework on the platform (which we hope :crossed_fingers: ), then we can reassess.