xla
[ragged-paged-attn] Use hidden states in kv cache and support any num_kv_head
#8851
Merged

[ragged-paged-attn] Use hidden states in kv cache and support any num_kv_head #8851

bythew3i
bythew3i Refactor kv cache to use hidden states
07d1cfe7
bythew3i Fix undefined variable
3a891781
bythew3i Fix kv cache shape in tests
ddbe8056
vanbasten23
vanbasten23 commented on 2025-03-18
vanbasten23
vanbasten23 commented on 2025-03-18
vanbasten23
vanbasten23 commented on 2025-03-18
vanbasten23
vanbasten23 commented on 2025-03-18
vanbasten23
vanbasten23 commented on 2025-03-18
vanbasten23
vanbasten23 commented on 2025-03-18
bythew3i Rename kv_model_dim to kv_hidden_size
add6be02
vanbasten23
vanbasten23 approved these changes on 2025-03-19
vanbasten23
bythew3i
vanbasten23 vanbasten23 merged 4190fc0e into master 230 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone