16:26:45
| * ryah | topic: http://propelml.org/ |
16:26:56
| <piscisaureus1> | indutny: yeah, about a year ago |
16:27:25
| <piscisaureus1> | ryah: ah thnks |
16:28:42
| <piscisaureus1> | indutny: but I think last time I saw you must have been at least 2 years ago. Your family was there too IIRC. Are you still living in Oakland? |
16:28:51
| <indutny> | I never did |
16:28:53
| <indutny> | :D |
16:29:11
| <indutny> | We lived in SF for 2-3 months |
16:29:13
| <indutny> | and then moved to NYC, where we stay now |
16:29:13
| <piscisaureus1> | oh i thought that you were (or were planning to) |
16:29:19
| <piscisaureus1> | ah that's right |
16:40:34
| <ryah> | piscisaureus1: https://github.com/propelml/propel/pull/432 |
16:40:48
| <ryah> | this should soon allow us to display dataset images in the docs with zero overhead |
16:45:52
| <piscisaureus1> | @ryah - approved it. I don't believe the zero overhead. |
16:46:02
| <piscisaureus1> | Base64 encoding images does not ring very efficient to me |
16:46:27
| <piscisaureus1> | But there is a lot to be said for the simplicity it adds. Canvas is hard to serialize. |
16:46:34
| <ryah> | piscisaureus1: well dataURLs are certainly way faster than re-rendering the image |
16:46:44
| <ryah> | we have to keep an eye on the size tho |
16:46:54
| <ryah> | i put an assert in there just to warn us |
16:47:24
| <ryah> | http://propelml.org/docs/#imread <-- currently nothing show |
16:47:43
| <piscisaureus1> | what if you added `du -hs build/website_render/*.js *.html *.css` to the end of the CI scripts |
16:47:54
| <piscisaureus1> | that helps us keep an eye on the size and its growth |
16:48:18
| <ryah> | does that work on windows? |
16:48:55
| <piscisaureus1> | I think git ships with du |
16:49:03
| <piscisaureus1> | and git is available on appveyor |
17:01:40
| <ryah> | im going to deploy the website |
17:03:53
| <ryah> | deployed |
17:07:35
| <ryah> | http://propelml.org/docs/#imread <--- extremely proud we can do this |
17:07:54
| <ryah> | doesn't look like much now... but wait til we have datasets with images statically rendered |
17:08:49
| <ryah> | we should really get https working |
17:11:45
| <ryah> | piscisaureus1: try this |
17:11:51
| <ryah> | import * as pr from "propel"; |
17:11:52
| <ryah> | let ds = pr.dataset("mnist/test").batch(10); |
17:11:52
| <ryah> | let { images } = await ds.next(); |
17:11:53
| <ryah> | pr.imshow(images.slice(0, 1).reshape([28, 28, 1])) |
17:12:01
| <ryah> | in the notebook |
17:14:29
| <piscisaureus1> | yeah nice |
17:14:31
| <piscisaureus1> | we need ssl |
17:14:43
| <piscisaureus1> | also some protection against running a notebook cell that's already running |
17:14:50
| <piscisaureus1> | but ssl first :) |
17:15:18
| <piscisaureus1> | the progress bar helps! |
17:15:27
| <ryah> | right? makes it look so much better |
17:16:01
| <piscisaureus1> | some indication that something is happening is really helpful |
17:26:33
| <ryah> | i've attached the formatting stuff to 3.3.0 milestone |
17:26:34
| <ryah> | https://github.com/propelml/propel/issues?q=is%3Aopen+is%3Aissue+milestone%3Av3.3.0 |
17:26:51
| <ryah> | once that and my caching stuff is landed, we can release |
17:26:55
| <ryah> | i should test CUDA now.. |
17:28:30
| * ryah | quit (Remote host closed the connection) |
17:29:05
| * ryah | joined |
17:32:33
| <indutny> | someone said SSL? |
17:32:35
| <indutny> | :D |
17:32:57
| <indutny> | if this is for propelml.org |
17:33:10
| <indutny> | I'd recommend https://caddy.community/ |
17:33:12
| <indutny> | wrong link |
17:33:14
| <indutny> | https://caddyserver.com/ |
17:34:17
| * ryah | quit (Ping timeout: 276 seconds) |
18:02:20
| <piscisaureus1> | we're hosting the website in an aws bucket |
18:26:11
| * piscisaureus1 | quit (Ping timeout: 268 seconds) |
19:08:46
| * ryah | joined |
19:13:50
| * qti3e | joined |
19:18:19
| * qti3e | quit (Ping timeout: 260 seconds) |
19:29:57
| * ryah | quit (Ping timeout: 240 seconds) |
19:59:03
| * piscisaureus | joined |
20:01:41
| * piscisaureus | quit (Client Quit) |