lucidrains / marge-pytorch Goto Github PK
View Code? Open in Web Editor NEWImplementation of Marge, Pre-training via Paraphrasing, in Pytorch
License: MIT License
Implementation of Marge, Pre-training via Paraphrasing, in Pytorch
License: MIT License
Can you please upload the data of train.dat
and its mask? Where we can download it?
nevermind, figured it out :)
Could you please explain the distillation loss?
It looks like you are doing a comparison between cross attention and doc_similarity?
def` distill_attn_loss(evi_dots, doc_similarities, mask = None, eps = 1e-5):
evi_dots = rearrange(evi_dots, 'b l h i n j -> b (l h i) n j')
if exists(mask):
mask = rearrange(mask, 'b n j -> b () n j')
evi_dots.masked_fill_(~mask, 0.)
denom = mask.expand_as(evi_dots).sum(dim = (1, -1))
evi_dots_mean = evi_dots.sum(dim = (1, -1)) / (denom + eps)
else:
evi_dots_mean = evi_dots.mean(dim = (1, -1))
normed_evi_dots = evi_dots_mean.softmax(dim = -1)
normed_evi_dots.detach_()
doc_similarities = doc_similarities.softmax(dim = -1).log()
loss = F.kl_div(doc_similarities, normed_evi_dots, reduction = 'batchmean')
return loss
Could you release a pre-trained model and provide some instructions on how to run it on the summarization dataset? Thanks.
What kind of search method are you using? do you use faiss library?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.