eagle3: add qwen3.5 4B 9B 35B-A3B support #21437
36330
wants to merge
18 commits into
ggml-org:master
from
36330:pr/eagle3-more
feat: add EAGLE3 speculative decoding support
8fac4b1c
fix eagle3 logits sync bug & remove ggml_set_sync()
ac5667dc
Merge branch 'master' into pr/18039
3e7f376b
eagle3 : improve naming
5a79c190
add eagle3 support for Qwen3 series models
c0d99e65
add eagle3 support for Qwen3 MoE models
71ba283a
eagle3: load lm_head from target model if not in draft model when con…
3da288d7
eagle3: make d2t mapping optional
13a9f31d
eagle3: add support for gpt-oss-120B eagle3
75883cde
eagle3: add support for RedHtAI eagle3 speculator series models
7b78bfa9
Merge branch 'master' into HEAD
7d4c2239
Merge branch 'master' into pr/18039
5e224bc1
eagle3: fix model convert issue
b3537924
eagle3: fix model convert code format
9fea2434
Merge branch 'master' into pr/18039
b8ab2cc5
eagle3: support --eagle3 in llama-cli
07e2c970
Merge branch 'master' into pr/18039
5bb2d50c
eagle3: add recurrent verification state support
57438283
Login to write a write a comment.
Login via GitHub