-
This is a major feature release.
Spec: https://github.com/MadcowD/ell/blob/cd64ab9bb0d3a09195fef7a32ef77ac5d7e6c912/docs/ramblings/evalspec.md
Ramblings: https://github.com/MadcowD/ell/blob/cd64ab9…
-
Your DG does not seem to have any object diagram, which makes the analysis and evaluation of the classes less comprehensive.
[original: nus-cs2113-AY2425S1/pe-interim#639] [original labels: seve…
-
Your work is amazing!
Can you also release the scripts of the Open-Vocabulary extension mentioned in the paper and evaluation codes for Open-Vocabulary segmentation? Many thanks!
-
Hi! Thank you for sharing your re-implementation!
I cannot seem to find any of your evaluation results, and I was wondering if you were able to achieve the same results (AUC) as the original HIPT pap…
-
Hi, thanks for the amazing work.
Can you please clarify how do you compute your error metrics, exactly?
Tab. V shows the relative error. Do you use the RPE-script from https://cvg.cit.tum.de/dat…
-
Hi, @luost26! Great work! I have a question regarding the evaluation.
I am comparing the evaluation of RDE-Net and RDE-Linear. Why are there fewer entries in the case of RDE-Linear? Where are they …
-
I had code in weave `v0.50.14` that successfully passed data into models through the `Evaluation` class. These objects are passed into the model `predict` method as dict-like objects, allowing usage l…
-
Cet issu c'est pour reflechir et proposer une solution pour la relation entre les deux tunnels : TD gaspi et evaluations gaspi
à faire : proposer une solution à l'équipe
-
This project demonstrates a solid grasp of both basic and advanced methods for solving the TSP, with detailed descriptions and thoughtful comparison of results. By implementing additional visualizatio…
-
Hi, you are doing a great job. I would like to reproduce your work, where can I find the relevant code for evaluating the generated text.
I run
``python test.py --length=50 --num_iterations=1 --tem…