DeepSpeed
Read rope_theta from rope_parameters in the Llama injection policy
#8341
Merged

Read rope_theta from rope_parameters in the Llama injection policy #8341

alanhuangyoo
alanhuangyoo Read rope_theta from rope_parameters in the Llama injection policy
ecda5f94
alanhuangyoo alanhuangyoo requested a review from hwchen2017 hwchen2017 19 days ago
alanhuangyoo alanhuangyoo requested a review from loadams loadams 19 days ago
alanhuangyoo alanhuangyoo requested a review from tjruwase tjruwase 19 days ago
alanhuangyoo alanhuangyoo requested a review from tohtana tohtana 19 days ago
alanhuangyoo
ebarkhordar
ebarkhordar commented on 2026-09-04
alanhuangyoo Pin the Llama rope_theta helper against a real LlamaConfig
3dd2716a
tohtana
tohtana requested changes on 2026-09-05
alanhuangyoo Refuse rotary variants kernel injection cannot serve
8447987c
alanhuangyoo alanhuangyoo force pushed from 16c9b4cf to 8447987c 10 days ago
alanhuangyoo
alanhuangyoo
alanhuangyoo Merge remote-tracking branch 'upstream/master' into fix/llama-injecti…
ee60682a
alanhuangyoo
tohtana
tohtana approved these changes on 2026-09-08
tohtana Merge branch 'master' into fix/llama-injection-rope-parameters
e8ba2410
tohtana tohtana enabled auto-merge 9 days ago
alanhuangyoo Apply yapf formatting to the rope_theta test
4c687f12
disabled auto-merge 9 days ago
Head branch was pushed to by a user without write access
tohtana tohtana enabled auto-merge 9 days ago
tohtana tohtana merged 6474bc55 into master 8 days ago
alanhuangyoo

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone