SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

[Styling] stylify using ruff (#27144)

Browse Source 
* try to stylify using ruff

* might need to remove these changes?

* use ruf format andruff check

* use isinstance instead of type comparision

* use # fmt: skip

* use # fmt: skip

* nits

* soem styling changes

* update ci job

* nits isinstance

* more files update

* nits

* more nits

* small nits

* check and format

* revert wrong changes

* actually use formatter instead of checker

* nits

* well docbuilder is overwriting this commit

* revert notebook changes

* try to nuke docbuilder

* style

* fix feature exrtaction test

* remve `indent-width = 4`

* fixup

* more nits

* update the ruff version that we use

* style

* nuke docbuilder styling

* leve the print for detected changes

* nits

* Remove file I/O

Co-authored-by: charliermarsh
 <charlie.r.marsh@gmail.com>

* style

* nits

* revert notebook changes

* Add # fmt skip when possible

* Add # fmt skip when possible

* Fix

* More `  # fmt: skip` usage

* More `  # fmt: skip` usage

* More `  # fmt: skip` usage

* NIts

* more fixes

* fix tapas

* Another way to skip

* Recommended way

* Fix two more fiels

* Remove asynch
Remove asynch

---------

Co-authored-by: charliermarsh <charlie.r.marsh@gmail.com>
This commit is contained in:
Arthur
2023-11-16 17:43:19 +01:00
committed by GitHub
parent acb5b4aff5
commit 651408a077
480 changed files with 867 additions and 1059 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
tests/models/nllb_moe/test_modeling_nllb_moe.py
View File

@@ -408,9 +408,7 @@ class NllbMoeModelIntegrationTests(unittest.TestCase):
with torch.no_grad():
output = model(**self.model_inputs)
# fmt: off
EXPECTED_LOGTIS = torch.Tensor([-0.3059, 0.0000, 9.3029, 0.6456, -0.9148, 1.7836, 0.6478, 0.9438, -0.5272, -0.6617, -1.2717, 0.4564, 0.1345, -0.2301, -1.0140, 1.1427, -1.5535, 0.1337, 0.2082, -0.8112, -0.3842, -0.3377, 0.1256, 0.6450, -0.0452, 0.0219, 1.4274, -0.4991, -0.2063, -0.4409,])
# fmt: on
EXPECTED_LOGTIS = torch.Tensor([-0.3059, 0.0000, 9.3029, 0.6456, -0.9148, 1.7836, 0.6478, 0.9438, -0.5272, -0.6617, -1.2717, 0.4564, 0.1345, -0.2301, -1.0140, 1.1427, -1.5535, 0.1337, 0.2082, -0.8112, -0.3842, -0.3377, 0.1256, 0.6450, -0.0452, 0.0219, 1.4274, -0.4991, -0.2063, -0.4409,]) # fmt: skip
torch.testing.assert_allclose(output.logits[1, 0, :30], EXPECTED_LOGTIS, rtol=6e-3, atol=9e-3)
@unittest.skip("This requires 300GB of RAM")
@@ -474,6 +472,7 @@ class NllbMoeRouterTest(unittest.TestCase):
Original implementation of the routers here:
"""
config = NllbMoeConfig(
num_experts=4,
hidden_size=32,
@@ -516,9 +515,7 @@ class NllbMoeRouterTest(unittest.TestCase):
masked_hidden_states[idx, token_indices] = torch.einsum("b,be->be", combining_weights, expert_output)
hidden_states = masked_hidden_states.sum(dim=0).reshape(self.batch_size, self.sequence_length, hidden_dim)
# fmt: off
EXPECTED_MEAN_FAIRSEQ_HIDDEN_STATES = torch.Tensor([[ 7.0340e-04, 2.7997e-03, -1.3351e-02, -7.6705e-03, -3.5089e-03,3.9773e-03, 7.4593e-03, 1.2566e-02, 3.5860e-03, -2.7448e-02,-1.3731e-02, -1.0534e-02, -1.3606e-02, -1.5048e-02, -2.8914e-03,-5.0371e-03, -1.3963e-03, 6.0076e-03, -1.1380e-02, -1.4620e-02, 5.2401e-03, 8.4660e-04, -1.5319e-03, -1.6735e-02, 1.1302e-02, 3.6119e-03, 4.6084e-03, -1.3458e-02, 7.7792e-05, 1.4312e-02, 4.9107e-03, -5.0936e-03], [-4.4538e-03, 3.1026e-03, 1.4121e-04, -4.8121e-03, -5.6279e-03, 7.2493e-03, 3.9769e-03, 1.1114e-02, -1.5666e-03, -2.3477e-02, 8.7268e-03, 1.3446e-02, -2.8845e-05, -1.7287e-02, 8.7619e-03, -4.5316e-03, -1.2164e-02, 5.7461e-03, -4.5861e-03, -9.3907e-03, 2.9808e-02, 8.9206e-04, -7.6232e-04, -1.4173e-02, 3.0208e-03, 1.5310e-02, 9.7717e-03, 3.1014e-03, 7.8042e-03, 8.0197e-03, 3.4784e-03, -7.1728e-03]])
# fmt: on
EXPECTED_MEAN_FAIRSEQ_HIDDEN_STATES = torch.Tensor([[ 7.0340e-04, 2.7997e-03, -1.3351e-02, -7.6705e-03, -3.5089e-03,3.9773e-03, 7.4593e-03, 1.2566e-02, 3.5860e-03, -2.7448e-02,-1.3731e-02, -1.0534e-02, -1.3606e-02, -1.5048e-02, -2.8914e-03,-5.0371e-03, -1.3963e-03, 6.0076e-03, -1.1380e-02, -1.4620e-02, 5.2401e-03, 8.4660e-04, -1.5319e-03, -1.6735e-02, 1.1302e-02, 3.6119e-03, 4.6084e-03, -1.3458e-02, 7.7792e-05, 1.4312e-02, 4.9107e-03, -5.0936e-03], [-4.4538e-03, 3.1026e-03, 1.4121e-04, -4.8121e-03, -5.6279e-03, 7.2493e-03, 3.9769e-03, 1.1114e-02, -1.5666e-03, -2.3477e-02, 8.7268e-03, 1.3446e-02, -2.8845e-05, -1.7287e-02, 8.7619e-03, -4.5316e-03, -1.2164e-02, 5.7461e-03, -4.5861e-03, -9.3907e-03, 2.9808e-02, 8.9206e-04, -7.6232e-04, -1.4173e-02, 3.0208e-03, 1.5310e-02, 9.7717e-03, 3.1014e-03, 7.8042e-03, 8.0197e-03, 3.4784e-03, -7.1728e-03]]) # fmt: skip
self.assertTrue(torch.allclose(hidden_states.mean(1), EXPECTED_MEAN_FAIRSEQ_HIDDEN_STATES, 1e-4))
def test_batch_prioritized_routing(self):