Updating blocking dummy tests by kdulla · Pull Request #1171 · quic/efficient-transformers

kdulla · 2026-07-10T07:23:07Z

Extend range of dummy tests to cover all models for which blocking is supported.
Added hkv and hq blocking tests to full test (these use the same forward as hqkv blocking so shouldn't need to be tested on dummy model level)
All updated tests can be ran with:
pytest tests/transformers/models/causal_lm_models/test_causal_lm_blocking_hqkv.py
Updates to hqkv forward to be compatible with manually passed hkv and hq blocking modes
Update to gpt-oss to properly downgrade q blocking to no blocking since gpt-oss does not support q blocking currently.

Signed-off-by: Kushal Dulla <kdulla@qti.qualcomm.com>

Signed-off-by: Kushal Dulla <quic_kdulla@quicinc.com>

Signed-off-by: Kushal Dulla <kdulla@qti.qualcomm.com>

kdulla and others added 3 commits July 7, 2026 13:36

updated blocking tests with dummy models

3d577b1

Signed-off-by: Kushal Dulla <kdulla@qti.qualcomm.com>

updated dummy tests and added downgrade to gpt-oss q blocking

1f98c56

Signed-off-by: Kushal Dulla <quic_kdulla@quicinc.com>

Fixed dummy test suite and hkv and hq blockign modes

9609bc4

Signed-off-by: Kushal Dulla <kdulla@qti.qualcomm.com>

kdulla requested review from ochougul and quic-rishinr July 10, 2026 07:23

kdulla self-assigned this Jul 10, 2026

kdulla added the qeff.blocking label Jul 10, 2026

reduced scope of dummy tests to relevant models and blocking modes

47c3c42

Signed-off-by: Kushal Dulla <kdulla@qti.qualcomm.com>

kdulla requested a review from vbaddi July 15, 2026 04:48

kdulla marked this pull request as draft July 24, 2026 08:42

Provide feedback