This issue is tracking the deprecation of the (1) configs (2) model checkpoints quantized with these configs.

What is deprecated:

• We added version 2 Int4WeightOnlyConfig in various PRs in Migrating from AffineQuantizedTensor + Layouts to new structure of tensor subclasses #2752 and switched the default version to 2 in Bump Int4WeightOnlyConfig version to 2 #2949, the version 1 config is now deprecated, please use version 2 config to quantize the model
• the quantized checkpoints quantized with version 1 config previously is deprecated as well, and we plan to remove the support to load these checkpoints after pytorch 2.11 release (around 9 months from now)

Timeline:
0.14.0: annouce deprecation for version 1 config
after all tensors are migrated: remove support for version 1 config
after pytorch 2.11 release: remove support for version 1 checkpoints