Home
Archive
About
menu
aPaperADay
aPaperADay
aPaperADay.
read a deep learning
paper a day
Archive
About
keyboard_arrow_up
10
-
12
chevron_left
chevron_right
5 UMLFiT Analysis
Today I look at the analysis of the the results of the UMLFiT paper.
2019-12-31
5 UMLFiT The "Chain Thaw"
The Fine-tune step is critically important in transfering knowledge to a specialized learned, boosted by Gradual Unfreezing.
2019-12-30
5 UMLFiT Getting Started
Fine tuning and adjustable learnign rates are key to success outlined in the UMLFiT paper.
2019-12-27
5 ULM-FiT Introduction
I get started on the ULMFiT paper, looking at the introduced ideas.
2019-12-26
Merry Christmas
May your Christmas be merry and bright. -John 1:9
2019-12-25
4 Finishing up Benchmarking
Finishing up the 90 page benchmarking results, nuanced results.
2019-12-24
4 Point to point BMs
In this section, we'll review the data transfers from one tile to another under diverse load conditions.
2019-12-21
4 Local Memory BMs
The latency and bandwidth are reviewed in ch 3.
2019-12-20
4 Experimental Setup
A quick look at the experimental setup used in the microbenchmarks.
2019-12-19
4 IPU BSP
The Bulk Synchronous Parallel Model is discussed.
2019-12-18
chevron_left
10/12
chevron_right