transformers
Add Deepseek-OCR-2 model
#45075
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
86
Changes
View On
GitHub
Add Deepseek-OCR-2 model
#45075
vasqu
merged 86 commits into
huggingface:main
from
thisisiron:add-deepseek_ocr2
feat: add DeepSeek-OCR-2 model registration and docs
e6bed84a
feat: add DeepseekOcr2ImageProcessor
7a1a4007
feat: add DeepseekOcr2Processor and refactor image processor tile_size
f3ec285d
feat: add script to convert DeepSeek-OCR-2 weights to Hugging Face fo…
95b0d574
feat: enhance DeepSeek-OCR-2 processing and inference test
b81bc0fb
refactor
bbf3f942
feat: add fast image processor
2e05944a
refactor
8f3270b4
test: add image processor tests for DeepseekOcr2
3ee14eba
fix: make background_color configurable
58b683eb
refactor: migrate image processors to pil/torchvision backend pattern
37104993
test: add processor tests for DeepseekOcr2
1bfa054f
fix: update __init__
1878af2c
chore: clean up unused imports and fix formatting
46ddaecc
feat: add configuration, modeling, and modular for DeepseekOcr2
8754d8df
fix: style fixes, update docs, and minor cleanups
fef7b5fa
fix: use @strict
5520c318
fix: register private models
fee7de05
docs: add usage example and expand DeepSeek-OCR-2 model doc
85312185
fix: add checkpoint to auto_docstring
f6fc20ba
fix: remove comment
b4bfbf53
fix: remove unused max_query
e7755774
thisisiron
changed the title
[WIP] Add Deepseek-OCR2 model
Add Deepseek-OCR-2 model
232 days ago
zucchini-nlp
requested a review
from
zucchini-nlp
227 days ago
fix: clean up DeepSeek-OCR2 modular
44482df2
Merge branch 'main' into add-deepseek_ocr2
4b1605a6
docs: update date
c6f5eaf2
refactor: inherit SamVisionEncoder
f05c2529
refactor: use create_causal_mask with or_mask_function
74ee9f3f
Merge branch 'main' into add-deepseek_ocr2
25c54542
Merge branch 'main' into add-deepseek_ocr2
33a61593
docs: update date
2f931aaf
zucchini-nlp
commented on 2026-04-02
fix: address PR review
d194f99b
refactor: use modular for image processor
c08b036f
refactor: restructure DeepSeek-OCR-2 config, image processor, and pro…
5ce50295
Merge branch 'main' into add-deepseek_ocr2
917d0864
fix: sync hidden_size and rms_norm_eps from encoder_config to vision_…
eec01bc6
Merge branch 'main' into add-deepseek_ocr2
7ef21416
refactor: remove comment
10639ec7
fix
1efd730c
fix: correct EncoderConfig docstring example
9c70392d
zucchini-nlp
commented on 2026-04-10
refactor: add PIL image processor to modular
b9c75c62
refactor: address review comments on config, processor, and model
1433b320
Merge branch 'main' into add-deepseek_ocr2
22392adf
fix: adjust processing tests for image token expansion
6b95f093
fix: move view_separator to correct device for model parallelism
55fc4aab
Merge branch 'main' into add-deepseek_ocr2
68b5d42e
fix: add DeepseekOcr2ImageProcessorPil to __all__
f85aabb2
fix: remove SDPA skip
c96106a0
test: skip offload/export tests
8dbfda57
zucchini-nlp
approved these changes on 2026-04-14
refactor: address review comments
8b391e63
Merge branch 'main' into add-deepseek_ocr2
0102cbf5
fix: remove in DeepseekOcr2Model
451fd53c
refactor: enforce explicit tokens in DeepseekOcr2Processor
0f73ee0e
Merge branch 'main' into add-deepseek_ocr2
50b210cc
refactor: inherit DeepseekOcr2ImageProcessorKwargs from GotOcr2ImageP…
f81b8b9f
refactor: remove unused image processing
d84deaec
zucchini-nlp
requested a review
from
Cyrilvallez
214 days ago
zucchini-nlp
requested a review
from
vasqu
214 days ago
vasqu
commented on 2026-04-21
fix: update docs
765a1562
Merge remote-tracking branch 'origin/main' into add-deepseek_ocr2
45aab0f1
fix: export sub-configs
cd728815
Update src/transformers/models/deepseek_ocr2/modular_deepseek_ocr2.py
54294aca
refactor: switch to mlp_layer_types pattern and clean up redundant ov…
9310aafc
refactor: derive projector dims from sub-configs
78fdc1cf
refactor: address comments on naming, decorators, and mask location
a73c8a4d
refactor: use VLMModelTester, fix processor text handling, and cache …
9389109f
Merge branch 'add-deepseek_ocr2' of https://github.com/thisisiron/tra…
dddd6851
fix: enable fullgraph compile
d9180fdc
refactor: move weight mapping to conversion_mapping.py
916c8b0d
fix: add tie_word_embeddings
9a9b7362
Merge branch 'main' into add-deepseek_ocr2
1d133011
vasqu
approved these changes on 2026-04-28
Merge branch 'main' into add-deepseek_ocr2
3d7c8845
refactor: address review comments
829f82b8
Merge branch 'main' into add-deepseek_ocr2
b7ae2eb0
vasqu
approved these changes on 2026-04-30
make ci compatible, passed without changes neat
0c18ada2
fixup imports
2ba5ede3
Merge branch 'main' into add-deepseek_ocr2
865faa95
switch vision encoder mask to block_sequence_ids
20035528
Merge branch 'main' into add-deepseek_ocr2
17ee7041
update docs
28fa3a52
vasqu
approved these changes on 2026-05-25
Merge branch 'main' into add-deepseek_ocr2
2772070d
update docs
077e99e0
Clean up DeepSeek-OCR-2 converted config
3c8fb527
update DeepSeek-OCR-2 checkpoint
36c883fc
Merge branch 'main' into add-deepseek_ocr2
9482d7e3
update docs and sync generated config
fd0e526c
vasqu
approved these changes on 2026-06-01
vasqu
added
New model
fix
b4069125
vasqu
enabled auto-merge
167 days ago
Merge branch 'main' into add-deepseek_ocr2
cf874952
ok
5b87e5ff
vasqu
enabled auto-merge
167 days ago
disabled auto-merge
167 days ago
Manually disabled by user
vasqu
enabled auto-merge
167 days ago
vasqu
merged
8e67c5e1
into main
167 days ago
Login to write a write a comment.
Login via GitHub
Reviewers
vasqu
zucchini-nlp
Cyrilvallez
Assignees
No one assigned
Labels
New model
Milestone
No milestone
Login to write a write a comment.
Login via GitHub