Rémi Ouazan remi-or

Kernel engineer at Hugging Face 🤗 also a transformers maintainer

remi-or / resize_comparaison.py

Created June 4, 2025 18:07

Manual resize comparaison

	from typing import Tuple
	import torch
	from torch import Tensor
	from torchvision.transforms.v2 import functional as F


	def reference_resize(image: Tensor, new_size: Tuple[int, int]) -> Tensor:
	image = F.resize(image, new_size, interpolation=F.InterpolationMode.BICUBIC, antialias=True)
	return image

remi-or / resize_compile_issue.py

Created June 2, 2025 16:26

MREx for a compile issue linked to resize

	import torch
	from torch import nn
	from torch.nn import functional as F
	from torchvision.transforms.v2 import functional as F
	import matplotlib.pyplot as plt


	INPUT_SIZE = (3, 41, 70)

remi-or / Augsup

Last active June 10, 2022 19:25

	def aug_test(self,
	imgs: List[Tensor],
	img_metas: List[dict],
	rescale: bool = False) -> Tensor:
	acc_boxes = np.zeros((0, 5))
	acc_score = np.zeros((0, self.roi_head.bbox_head.num_classes))
	for img, img_meta in zip(imgs, img_metas):
	for label, dets in enumerate(self.simple_test(img, img_meta, None, rescale)[0]):
	boxes, scores = dets[:, :-1], dets[:, -1]
	acc_boxes = np.vstack((acc_boxes, boxes))

remi-or / FCOSRside.ipynb

Created May 18, 2022 19:25

Sorry, something went wrong. Reload?

Sorry, we cannot display this file.

Sorry, this file is invalid so it cannot be displayed.

remi-or / distillator.py

Created January 19, 2022 22:16

	## Imports
	from typing import Tuple
	import torch
	from torch import Module, Tensor
	from transformers.models.roberta.modeling_roberta import RobertaPreTrainedModel, RobertaConfig, RobertaModel, RobertaEncoder
	from torch.nn import CrossEntropyLoss, CosineEmbeddingLoss



	## Function

remi-or / distillation_loss.py

Created January 18, 2022 17:55

	import torch
	from torch.nn import CrossEntropyLoss, CosineEmbeddingLoss

	def distillation_loss(
	teacher_logits : Tensor,
	student_logits : Tensor,
	labels : Tensor,
	temperature : float = 1.0,
	) -> Tensor:
	"""

remi-or / get_logits.py

Created January 18, 2022 17:36

	from torch import Tensor

	def get_logits(
	model : RobertaPreTrainedModel,
	input_ids : Tensor,
	attention_mask : Tensor,
	) -> Tensor:
	"""
	Given a RoBERTa (model) for classification and the couple of (input_ids) and (attention_mask),
	returns the logits corresponding to the prediction.

remi-or / distill_roberta_weights.py

Created January 17, 2022 18:34

	from transformers.models.roberta.modeling_roberta import RobertaEncoder, RobertaModel
	from torch.nn import Module

	def distill_roberta_weights(
	teacher : Module,
	student : Module,
	) -> None:
	"""
	Recursively copies the weights of the (teacher) to the (student).
	This function is meant to be first called on a RobertaFor... model, but is then called on every children of that model recursively.

remi-or / distill_roberta.py

Last active January 17, 2022 18:32

	from transformers.models.roberta.modeling_roberta import RobertaPreTrainedModel, RobertaConfig

	def distill_roberta(
	teacher_model : RobertaPreTrainedModel,
	) -> RobertaPreTrainedModel:
	"""
	Distilates a RoBERTa (teacher_model) like would DistilBERT for a BERT model.
	The student model has the same configuration, except for the number of hidden layers, which is // by 2.
	The student layers are initilized by copying one out of two layers of the teacher, starting with layer 0.
	The head of the teacher is also copied.

remi-or / roberta_recursive_peek.py

Last active January 17, 2022 18:22

	from typing import Any
	from transformers import AutoModelForMaskedLM

	roberta = AutoModelForMaskedLM.from_pretrained("roberta-large")

	def visualize_children(
	object : Any,
	level : int = 0,
	) -> None:
	"""