transformers
Add deepseek 3.2 exp
#41251
Merged

Add deepseek 3.2 exp #41251

vasqu merged 31 commits into main from add-deepseek-exp
ArthurZucker
HuggingFaceDocBuilderDev
HandH1998
ArthurZucker
ArthurZucker
ArthurZucker ArthurZucker marked this pull request as ready for review 263 days ago
nfywsh
yunkchen
nfywsh
yunkchen
nfywsh
bmtwl
jyliu24
michaelroyzen
freedom-cui
michaelroyzen
ArthurZucker
michaelroyzen
RissyRan
ArthurZucker
ArthurZucker
shuningjin
michaelroyzen
michaelroyzen
ArthurZucker
ArthurZucker
leideng
ArthurZucker ArthurZucker force pushed from a2ca5296 to 974f3652 60 days ago
ArthurZucker ArthurZucker force pushed from 974f3652 to 3a26469f 60 days ago
ArthurZucker Add DeepSeek V3.2-Exp (DeepSeek Sparse Attention)
e715cfe4
ArthurZucker ArthurZucker force pushed from 3a26469f to e715cfe4 59 days ago
ArthurZucker Indexer returns the additive index mask directly (drop unused top-k i…
5f364a8b
ArthurZucker Make deepseek_v32 a lean DeepSeek-V3 + DSA model; glm_moe_dsa inherit…
8f1a76fd
ArthurZucker Address review: matmul-only indexer, explicit attention forward, impo…
4c2576e8
ArthurZucker
ArthurZucker commented on 2026-05-11
ArthurZucker Revert DynamicCache __iter__ / ddp churn for indexer keys
e89da2af
ArthurZucker Docs: drop the GlmMoeDsa TIP and contributor line per review
e92bdccb
michaelroyzen
ArthurZucker Gate DSA mask materialization to eager/SDPA; indexer returns top-k in…
2803e9c1
ArthurZucker Simplify deepseek_v32 checkpoint conversion to the qwen2_moe base
27056a3b
ArthurZucker
ArthurZucker
molbap
molbap commented on 2026-06-09
ArthurZucker Fix CI: config-attributes check, indexer fp32 dtype + causality
5d22edc6
ArthurZucker Make first_k_dense_replace configurable and fix tiny-config MoE test …
0366c779
ArthurZucker more efficient rotary pos emb
2499d54f
ArthurZucker manual cleanups
caa2bdfa
ArthurZucker up
b5ec1697
ArthurZucker updates
15df809d
ArthurZucker nits
c7d00493
ArthurZucker nits
b56feb33
ArthurZucker nit
612e6f5b
ArthurZucker nits
f700d25e
ArthurZucker remove the overwrite with a post_init fix?
1a4be191
ArthurZucker fix tests
fa90aaa1
ArthurZucker long context test
bc9190ed
ArthurZucker Merge branch 'add-deepseek-exp' of github.com:huggingface/transformer…
17414cef
ArthurZucker Tidy deepseek_v32 config comments
ef4e5e8b
ArthurZucker Merge remote-tracking branch 'origin/main' into add-deepseek-exp
21cf984a
ArthurZucker add padded test and start styling
c0e5f2d8
ArthurZucker nit
806e5020
ArthurZucker ArthurZucker added New model
ArthurZucker Merge branch 'add-deepseek-exp' of github.com:huggingface/transformer…
f1e4ebff
ArthurZucker
ArthurZucker Apply ruff format to deepseek_v32 / glm_moe_dsa modular files
bc3e5887
ArthurZucker Merge branch 'add-deepseek-exp' of github.com:huggingface/transformer…
7f09fcd4
github-actions
ArthurZucker Add publish/contribute dates to deepseek_v32 model card
87c4504e
github-actions
vasqu Merge branch 'main' into add-deepseek-exp
42eec216
vasqu vasqu merged 513146c1 into main 58 days ago
vasqu vasqu deleted the add-deepseek-exp branch 58 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone