search

huggingface / optimum-tpu

Google TPU optimizations for transformers models
Apache License 2.0
76 stars 19 forks source link

Fix instruct models UI issue #78

Closed tengomucho closed 4 months ago

tengomucho commented 4 months ago

What does this PR do?

The logits process is relaxed so it allows top_k =1 and top_p=1 even when do_sample=True. These kind of requests was received when testing instruct models such as mistralai/Mistral-7B-Instruct-v0.3 with the inference endpoints web ui. This should fix #76 and #77. Version is increased, so I can release that and use the new version with the fix in Inference Endpoints.

HuggingFaceDocBuilderDev commented 4 months ago

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.