😶‍🌫️

Remy Liu RemyLau

😶‍🌫️

Post-doc fellow in the Ray and Stephanie Lane Computational Biology Department at CMU

RemyLau / parse_gtf.py

Last active February 4, 2023 16:09

Parse GTF file into a dataframe by extracting the attributes from the last (attr) column

	import pandas as pd

	# GTF files, e.g. https://ftp.ensembl.org/pub/current_gtf/homo_sapiens/Homo_sapiens.GRCh38.108.gtf.gz
	# README for GTF: https://ftp.ensembl.org/pub/current_gtf/homo_sapiens/README
	raw_df = pd.read_csv("some_file.gtf", sep="\t", comment="#", header=None)

	# Parse attributes from the last column
	# Example value: 'gene_id "ENSG00000160072"; gene_version "20"; gene_name "ATAD3B"; gene_source "ensembl_havana"; gene_biotype "protein_coding";'
	# First convert each to a dict: {'gene_id': 'ENSG00000160072', 'gene_version': '20', 'gene_name': 'ATAD3B', 'gene_source': 'ensembl_havana', 'gene_biotype': 'protein_coding'}
	# Then combine the list of dict into a dataframe

RemyLau / pytorch_loss_brief.py

Last active March 20, 2022 20:11

Understanding the difference between cross entropy and negative log-likelihood loss as implemented in PyTorch (brief version)

	import torch

	torch.manual_seed(0)

	# Binary setting ##############################################################
	print(f"{'Setting up binary case':-^80}")
	z = torch.randn(5)
	yhat = torch.sigmoid(z)
	y = torch.Tensor([0, 1, 1, 0, 1])
	print(f"{z=}\n{yhat=}\n{y=}\n{'':-^80}")

RemyLau / pytorch_loss.py

Last active March 12, 2026 05:33

Understanding the difference between cross entropy and negative log-likelihood loss as implemented in PyTorch

	import torch
	import torch.nn.functional as F

	torch.manual_seed(0)

	# Binary setting ##############################################################
	print(f"{'Setting up binary case':-^80}")
	z = torch.randn(5)
	yhat = torch.sigmoid(z)
	y = torch.Tensor([0, 1, 1, 0, 1])