llama.cpp
Allow pooled embeddings on any model
#7477
Merged

Allow pooled embeddings on any model #7477

iamlemec
github-actions github-actions added examples
github-actions
compilade
compilade commented on 2024-05-22
compilade compilade added enhancement
compilade compilade added embeddings
compilade compilade added Review Complexity : Medium
compilade
compilade commented on 2024-06-01
ggerganov
ggerganov commented on 2024-06-04
ggerganov
ggerganov approved these changes on 2024-06-07
ggerganov ggerganov requested a review from compilade compilade 1 year ago
ngxson
iamlemec create append_pooling operation; allow to specify attention_type; add…
01057149
iamlemec find result_norm/result_embd tensors properly; update output allocati…
1756c4b5
iamlemec only use embd output for pooling_type NONE
7c37ae9d
iamlemec get rid of old causal_attn accessor
d4e6972f
iamlemec take out attention_type; add in llama_set_embeddings
8093253b
iamlemec iamlemec force pushed to 8093253b 1 year ago
iamlemec
compilade
compilade commented on 2024-06-14
iamlemec bypass logits when doing non-NONE pooling
5cc7b453
iamlemec
compilade
compilade approved these changes on 2024-06-20
ngxson
ngxson commented on 2024-06-20
ggerganov ggerganov merged 80ea089d into master 1 year ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone