Add raft_large weights fined-tuned on Kitti

[NicolasHug]

towards #4644

This PR adds pre-trained weights for raft-large, fine-tuned on Kitti.

I'm publishing both our weights and the original weights. The Kitti authors kindly allowed us to submit our code for evaluation on Kitti-test.

Our f1-epe on Kitti-test is 5.19 wheras the original is 5.10. We're a tiny bit higher (i.e. worse), but still significantly better than the other baselines that are compared against in the paper (next best one is 6.10).

There are submission restrictions, so to keep consistency with the C_T and C_T_SKHT weights, I just submitted the model that we already used in both (after kitti-specific fine-tuning). In other words these weights are litterally the current C_T_SKHT_V with the last fine-tuning step on Kitti.

Evaluated on kitti train:

(raft) ➜  vision git:(raft_kitti_weights) ✗ torchrun --nproc_per_node 8 --nnodes 1 references/optical_flow/train.py --val-dataset kitti --batch-size 10 --model raft_large --dataset-root ../downloads --weights Raft_Large_Weights.C_T_SKHT_K_V1

Kitti val epe: 0.6262	1px: 0.8732	3px: 0.9696	5px: 0.9859	per_image_epe: 0.6303	f1: 1.4712

(raft) ➜  vision git:(raft_kitti_weights) ✗ torchrun --nproc_per_node 8 --nnodes 1 references/optical_flow/train.py --val-dataset kitti --batch-size 10 --model raft_large --dataset-root ../downloads --weights Raft_Large_Weights.C_T_SKHT_K_V2

Kitti val epe: 0.6283	1px: 0.8752	3px: 0.9689	5px: 0.9852	per_image_epe: 0.6347	f1: 1.5217

kitti test submission:

cc @datumbox

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 2c14700 (more details on the Dr. CI page):

---

• 2/2 failures introduced in this PR

---

🕵️ 2 new failures recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

unittest_linux_cpu_py3.8 (1/2)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun)

/root/project/torchvision/io/video.py:399: Runt...log: [mov,mp4,m4a,3gp,3g2,mj2] moov atom not found

test/test_internet.py::TestDatasetUtils::test_download_url_dont_exist
  /root/project/env/lib/python3.8/site-packages/_pytest/fixtures.py:928: ResourceWarning: unclosed <ssl.SSLSocket [closed] fd=13, family=AddressFamily.AF_INET, type=SocketKind.SOCK_STREAM, proto=6>
    finalizer = functools.partial(_teardown_yield_fixture, fixturefunc, generator)

test/test_internet.py::TestDatasetUtils::test_download_url_dont_exist
  /root/project/env/lib/python3.8/site-packages/_pytest/fixtures.py:928: ResourceWarning: unclosed <ssl.SSLSocket [closed] fd=14, family=AddressFamily.AF_INET, type=SocketKind.SOCK_STREAM, proto=6>
    finalizer = functools.partial(_teardown_yield_fixture, fixturefunc, generator)

test/test_io.py::TestVideo::test_read_video_timestamps_corrupted_file
  /root/project/torchvision/io/video.py:399: RuntimeWarning: Failed to open container for /tmp/tmpd5lueftl.mp4; Caught error: [Errno 1094995529] Invalid data found when processing input: '/tmp/tmpd5lueftl.mp4'; last error log: [mov,mp4,m4a,3gp,3g2,mj2] moov atom not found
    warnings.warn(msg, RuntimeWarning)

test/test_models.py::test_memory_efficient_densenet[densenet121]
test/test_models.py::test_memory_efficient_densenet[densenet169]
test/test_models.py::test_memory_efficient_densenet[densenet201]
test/test_models.py::test_memory_efficient_densenet[densenet161]
  /root/project/env/lib/python3.8/site-packages/torch/autocast_mode.py:162: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
    warnings.warn('User provided device_type of \'cuda\', but CUDA is not available. Disabling')

test/test_models.py::test_inception_v3_eval

unittest_linux_cpu_py3.7 (2/2)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun)

/root/project/torchvision/io/video.py:399: Runt...log: [mov,mp4,m4a,3gp,3g2,mj2] moov atom not found

test/test_internet.py::TestDatasetUtils::test_download_url_dispatch_download_from_google_drive
  /root/project/env/lib/python3.7/unittest/mock.py:1966: ResourceWarning: unclosed <ssl.SSLSocket [closed] fd=13, family=AddressFamily.AF_INET, type=SocketKind.SOCK_STREAM, proto=6>
    self.name = name

test/test_internet.py::TestDatasetUtils::test_download_url_dispatch_download_from_google_drive
  /root/project/env/lib/python3.7/unittest/mock.py:1966: ResourceWarning: unclosed <ssl.SSLSocket [closed] fd=14, family=AddressFamily.AF_INET, type=SocketKind.SOCK_STREAM, proto=6>
    self.name = name

test/test_io.py::TestVideo::test_read_video_timestamps_corrupted_file
  /root/project/torchvision/io/video.py:399: RuntimeWarning: Failed to open container for /tmp/tmp34p85xv2.mp4; Caught error: [Errno 1094995529] Invalid data found when processing input: '/tmp/tmp34p85xv2.mp4'; last error log: [mov,mp4,m4a,3gp,3g2,mj2] moov atom not found
    warnings.warn(msg, RuntimeWarning)

test/test_models.py::test_memory_efficient_densenet[densenet121]
test/test_models.py::test_memory_efficient_densenet[densenet169]
test/test_models.py::test_memory_efficient_densenet[densenet201]
test/test_models.py::test_memory_efficient_densenet[densenet161]
  /root/project/env/lib/python3.7/site-packages/torch/autocast_mode.py:162: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling
    warnings.warn('User provided device_type of \'cuda\', but CUDA is not available. Disabling')

test/test_models.py::test_inception_v3_eval

---

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

[datumbox]

LGTM, thanks!