Code

Add support for Apple's Depth-Pro#34583

Code

Open

geetu040 wants to merge 73 commits intohuggingface:mainhuggingface/transformers:mainfrom geetu040:depth-progeetu040/transformers:depth-proCopy head branch name to clipboard

+3,696

Commits

Commits on Nov 3, 2024

• implement config and model building blocks

  geetu040

  committed
  2986dc2

Commits on Nov 9, 2024

• refactor model architechture

  geetu040

  committed
  1728a2f

Commits on Nov 12, 2024

• update model outputs

  geetu040

  committed
  11ce50c

Commits on Nov 16, 2024

• update init param to include use_fov_model

  geetu040

  committed
  27e9593

• update param name in config

  geetu040

  committed
  e74a7f5

• fix hidden_states and attentions outputs for fov

  geetu040

  committed
  8c2460b

• sort config

  geetu040

  committed
  55f6ed3

• complete minor todos

  geetu040

  committed
  b25dffb

• update patching

  geetu040

  committed
  c225deb

• update config for encoder

  geetu040

  committed
  176932d

• fix config

  geetu040

  committed
  dcec522

• use correct defaults in config

  geetu040

  committed
  0384d2f

Commits on Nov 17, 2024

• update merge for compatibility with different image size

  geetu040

  committed
  85e4f86

Commits on Nov 21, 2024

• restructure encoder for custom configuration

  geetu040

  committed
  00e4aa3

• make fov model compatible with custom config

  geetu040

  committed
  6be242c

• replace word "decoder" with "fusion"

  geetu040

  committed
  0189108

Commits on Nov 24, 2024

• weight conversion script

  geetu040

  committed
  7614e1a

Commits on Nov 25, 2024

• fix fov squeeze

  geetu040

  committed
  7d323ce

• update conversion script (without test)

  geetu040

  committed
  6aaa59e

• upload ruff image processing

  geetu040

  committed
  263b773

Commits on Nov 26, 2024

• create fast image processing

  geetu040

  committed
  17e5487

• use torch interpolation for image processing

  geetu040

  committed
  a8dd704

• complete post_process_depth_estimation

  geetu040

  committed
  261bbaf

• config: fix imports and sort args

  geetu040

  committed
  a4b3556

• apply inference in weight conversion

  geetu040

  committed
  f13c632

Commits on Nov 27, 2024

• use mllama script instead for weight conversion

  geetu040

  committed
  387ddd8

• clean weight conversion script

  geetu040

  committed
  9b67f9d

• add depth-pro status in other files

  geetu040

  committed
  617c872

• fill docstring in config

  geetu040

  committed
  6e1c512

• formatting

  geetu040

  committed
  12ee607

• more formatting

  geetu040

  committed
  d0a8733

• formatting with ruff

  geetu040

  committed
  e6b385a

• formatting with style

  geetu040

  committed
  267e50f

• fix copied classes

  geetu040

  committed
  a1ec997

• add examples; update weight convert script

  geetu040

  committed
  3c656f2

Commits on Nov 29, 2024

• fix using check_table.py and isort

  geetu040

  committed
  f6f6d3d

• fix config docstring

  geetu040

  committed
  b4575d0

• add depth pro to sdpa docs

  geetu040

  committed
  c8d8a9e

• undo unintentional changes in configuration_gemma.py

  geetu040

  authored
  77873de

Commits on Nov 30, 2024

• minor fixes

  geetu040

  committed
  5f2378d

• test image processing

  geetu040

  committed
  d51d0b1

Commits on Dec 2, 2024

• fixes and tests

  geetu040

  committed
  082b055

• more fixes

  geetu040

  committed
  16a3917

Commits on Dec 3, 2024

• use output states from image_encoder instead

  geetu040

  committed
  2408ec5

Commits on Dec 4, 2024

• Revert "use output states from image_encoder instead"

  Show description for be0c2a3
  geetu040

  committed
  be0c2a3

• make embeddings dynamic

  geetu040

  committed
  efed39f

• reshape output hidden states and attentions as part of computation graph

  geetu040

  committed
  c3b14fb

• fix ruff formating

  geetu040

  committed
  7cf2485

• fix docstring failure

  geetu040

  committed
  0aa451d

• use num_fov_head_layers in tests

  geetu040

  committed
  160afbf

• update doc

  geetu040

  committed
  9d2be26

• check consistency with config

  geetu040

  committed
  e208459

• ruff formatting

  geetu040

  committed
  0415722

• merge branch main

  geetu040

  committed
  402eedf

Commits on Dec 5, 2024

• update test case

  geetu040

  committed
  f4e7404

• fix ruff formatting

  geetu040

  committed
  2c1cc10

• Merge branch 'main' into depth-pro

  geetu040

  committed
  4d94396

Commits on Dec 6, 2024

• add tests for fov

  geetu040

  committed
  871b80d

• use interpolation in postprocess

  geetu040

  committed
  0ff0655

• run and fix slow tests locally

  geetu040

  committed
  befa6cd

• Merge branch 'main' into depth-pro

  geetu040

  committed
  db16fe6

Commits on Dec 12, 2024

• use scaled_images_features for image and fov encoder

  geetu040

  committed
  99ac5e8

• return fused_hidden_states in fusion stage

  geetu040

  committed
  ebb62dd

• fix example

  geetu040

  committed
  46c88e8

• fix ruff

  geetu040

  committed
  2431358

• Merge branch 'main' into depth-pro

  geetu040

  committed
  fd38841

Commits on Dec 21, 2024

• fix copyright license for all files

  geetu040

  committed
  d9d3a49

• add __all__ for each file

  geetu040

  committed
  8f4c61f

• minor fixes

  Show description for 8960535
  geetu040

  committed
  8960535

• return list in post_process_depth_estimation

  geetu040

  committed
  1ac1b84

• minor fixes

  Show description for 27bff69
  geetu040

  committed
  27bff69

• fix "ruff check"

  geetu040

  committed
  a69b5af

• update upsample and projection

  geetu040

  committed
  365a71d