Andrej Karpathy
|
d7a303b410
|
add thumbnails for papers, which apparently ppl like
|
2022-02-13 18:30:14 -08:00 |
|
Andrej Karpathy
|
759f7e73e6
|
fix bug in pagination, clean up the approach a bit more
|
2021-12-18 23:25:45 -08:00 |
|
Andrej Karpathy
|
c3cb157c9f
|
first version of pagination w00t w00t! it's a bit hacky i think, should be possible to improve this code and make it smaller and cleaner and etc.
|
2021-12-18 21:14:44 -08:00 |
|
Andrej Karpathy
|
4210a73d15
|
remove recommendation by default on / endpoint. this was causing some unintended confusion, e.g. the search function was using time_filter 7 and giving bad results from the main page. also getting recommendations takes time and is slowish, so only do it on actual demand
|
2021-12-18 20:21:04 -08:00 |
|
Andrej Karpathy
|
cd2a88f375
|
when a user is logged in and hits the default endpoint '/' show recommendations over the last week instead of just totally random recent papers, by default
|
2021-12-10 09:15:29 -08:00 |
|
Andrej Karpathy
|
33b2b018ab
|
kind of big changes here: we can now inspect & see the most positive and negative words for a trained svm, to help tune the value C. then there is also the UI for setting value C in the SVM. Finally the value of C I adjusted to default to 0.01 (was 0.1 before) because the results and the weights look more sensible based on manual inspection. We need some dataset of people libraries in order to potentially cross-validate a good value C automatically. For now there are not enough active users of the site that such a thing could be attempted and succeed. Doing my best now just by eyeballing
|
2021-12-08 00:17:37 -08:00 |
|
Andrej Karpathy
|
6be574a0fe
|
hah don't crash if a user submits an invalid query, better to return empty results. in the future we may want a helpful error message too
|
2021-12-07 22:56:09 -08:00 |
|
Andrej Karpathy
|
4ec9a75293
|
add a secret GET argument svm_c that changes the C value in the SVM. no UI for this yet :)
|
2021-12-07 22:52:13 -08:00 |
|
Andrej Karpathy
|
fe96775e6e
|
fix issues and improve logic of removing tags
|
2021-11-30 22:57:59 -08:00 |
|
Andrej Karpathy
|
9c00c8dacd
|
disallow the special tag null, which can result if user escapes the prompt
|
2021-11-30 22:51:16 -08:00 |
|
Andrej Karpathy
|
8947dd3fc5
|
tweak search scoring for better results, i think
|
2021-11-27 18:09:17 -08:00 |
|
Andrej Karpathy
|
5990a938a4
|
add few more stats
|
2021-11-27 14:47:28 -08:00 |
|
Andrej Karpathy
|
a169f0fa61
|
adding an about page as well
|
2021-11-27 11:47:17 -08:00 |
|
Andrej Karpathy
|
49995465df
|
add ability to store an email for a user for recommendations
|
2021-11-27 11:04:36 -08:00 |
|
Andrej Karpathy
|
970a9e24cf
|
maintain last active state for users
|
2021-11-27 09:37:02 -08:00 |
|
Andrej Karpathy
|
47fba66712
|
add a new stats page
|
2021-11-26 21:57:48 -08:00 |
|
Andrej Karpathy
|
c5b5c7833c
|
and i think that's it, we now support user accounts (lite)git commit -m 'and i think that\'s it, we now support user accounts litegit status sweet.'! sweet.
|
2021-11-26 16:38:36 -08:00 |
|
Andrej Karpathy
|
54157c6e9c
|
ok here we go, the beginning of user profiles. for now we assume there is a single hard-coded user root
|
2021-11-26 11:07:16 -08:00 |
|
Andrej Karpathy
|
3e2d1248d8
|
speed up time filtering by 20X by loading all of metas into memory once instead of querying it item by item in a massive loop
|
2021-11-26 10:22:52 -08:00 |
|
Andrej Karpathy
|
3d5bbb9851
|
optimization: read only metas if that's all that we need to complete the request. that said, i expected this to run much faster and it doesn't. i think the way i'm using sqlites here is not exactly efficient or appropriate and i have to re-think the database schema. e.g. getting random papers over the last 7 days should be almost instantaneous, but currently this request takes like 3 seconds. have to better understand why this is happening
|
2021-11-26 09:51:48 -08:00 |
|
Andrej Karpathy
|
b283c85c72
|
this was bothering me
|
2021-11-26 09:36:09 -08:00 |
|
Andrej Karpathy
|
8c74150337
|
eliminate a whole spurious endpoint of /search and remove code duplication
|
2021-11-26 09:31:06 -08:00 |
|
Andrej Karpathy
|
470bd00563
|
refactor and clean up the code in preparation for some consolidation, coming up
|
2021-11-26 09:10:35 -08:00 |
|
Andrej Karpathy
|
77279e1777
|
sequester all file sytem IO ops only to db.py, so it's not total chaos
|
2021-11-25 13:28:04 -08:00 |
|
Andrej Karpathy
|
cf1bef6f53
|
big new feature: ability to inspect any paper to see the raw tfidf tokens and their weights that summarize the paper, and which powers the SVM recommendation engine. basically a bit of a debugging / insight feature, but a really good sanity check that papers are being properly represented
|
2021-11-21 20:51:01 -08:00 |
|
Andrej Karpathy
|
8be24dae43
|
allow for full sorting, slice and dicing with all options in the UI, and the previous shortcuts. now have a lot more flexibility to slice and dice as desired
|
2021-11-16 21:39:43 -08:00 |
|
Andrej Karpathy
|
abb83effed
|
add ability to remove a tag from a paper as well
|
2021-11-14 21:49:34 -08:00 |
|
Andrej Karpathy
|
f02766be2c
|
oops need to open for writing if changing the tags dict
|
2021-11-12 21:23:22 -08:00 |
|
Andrej Karpathy
|
13a1d5ff48
|
sequester gross details about database instantiation in the filesystem away from the scripts
|
2021-11-12 21:12:09 -08:00 |
|
Andrej Karpathy
|
194b7f4b22
|
first leet codes
|
2021-11-12 20:40:19 -08:00 |
|