[quant][fix] Support quantization of ops where input is quantizable (#49027)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/49027
For cat followed by linear since the output of cat is not quanitzed, we didnt quantize the linear
This checks the uses of the cat op to insert observers
Test Plan:
python test/test_quantization.py TestQuantizeJitOps.test_cat_linear
Imported from OSS
Reviewed By: jerryzh168
Differential Revision: D25403412
fbshipit-source-id: 5875db259bf75f08ce672ce341a67005ed2f8a04