65 commits

Author	SHA1	Message	Date
Starbeamrainbowlabs	4e4d42a281	LossDice: add comment	2022-12-12 18:34:20 +00:00
Starbeamrainbowlabs	449bc425a7	LossDice: explicitly cast inputs to float32	2022-12-12 17:20:32 +00:00
Starbeamrainbowlabs	bcd2f1251e	LossDice: Do 1 - thing instead of -thing	2022-12-09 19:41:32 +00:00
Starbeamrainbowlabs	0129c35a35	LossDice: remove weird K.* functions	2022-12-09 19:06:26 +00:00
Starbeamrainbowlabs	659fc97fd4	fix crash	2022-12-09 18:39:27 +00:00
Starbeamrainbowlabs	649c262960	mono: switch loss from crossentropy to dice	2022-12-09 18:13:37 +00:00
Starbeamrainbowlabs	2a1772a211	confvnext_intrevse: add shallow	2022-12-08 19:10:12 +00:00
Starbeamrainbowlabs	a3c9416cf0	LossCrossentropy: don't sum	2022-12-08 16:57:11 +00:00
Starbeamrainbowlabs	f92b2b3472	according to the equation it looks like it's 2	2022-12-02 17:22:46 +00:00
Starbeamrainbowlabs	cad82cd1bc	CBAM: unsure if it's 1 ro 3 dense ayers in the shared mlp	2022-12-02 17:21:13 +00:00
Starbeamrainbowlabs	62f6a993bb	implement CBAM, but it's UNTESTED ... Convolutional Block Attention Module.	2022-12-02 17:17:45 +00:00
Starbeamrainbowlabs	01101ad30b	losscrossentropy: return the reduced value * facepalm *	2022-11-29 19:07:08 +00:00
Starbeamrainbowlabs	37f196a785	LossCrossentropy: add kwargs	2022-11-29 15:40:35 +00:00
Starbeamrainbowlabs	dba6cbffcd	WHY. * facepalms *	2022-11-28 19:33:42 +00:00
Starbeamrainbowlabs	57b8eb93fb	fixup	2022-11-28 19:09:35 +00:00
Starbeamrainbowlabs	6640a41bb7	almost got it....? it's not what I expected....!	2022-11-28 19:08:50 +00:00
Starbeamrainbowlabs	f48473b703	fixup	2022-11-28 19:00:11 +00:00
Starbeamrainbowlabs	f6feb125e3	this iss ome serious debugging. ... This commit will produce an extremely large volume of output.	2022-11-28 18:57:41 +00:00
Starbeamrainbowlabs	09f81b0746	train_mono: debug ... this commit will generate a large amount of debug output.	2022-11-28 16:46:17 +00:00
Starbeamrainbowlabs	f39e4ade70	LayerConvNextGamma: fix config serialisation bug ... .....this is unlikely to be the problem as this bug is in an unused code path.	2022-11-25 21:16:31 +00:00
Starbeamrainbowlabs	527b34942d	convnext_inverse: kernel_size 4→2	2022-11-11 19:29:37 +00:00
Starbeamrainbowlabs	44ad51f483	CallbackNBatchCsv: bugfix .sort() → sorted()	2022-11-04 16:40:21 +00:00
Starbeamrainbowlabs	1375201c5f	CallbackNBatchCsv: open_handle mode	2022-11-03 18:29:00 +00:00
Starbeamrainbowlabs	5f8d6dc6ea	Add metrics every 64 batches ... this is important, because with large batches it can be difficult to tell what's happening inside each epoch.	2022-10-31 19:26:10 +00:00
Starbeamrainbowlabs	172cf9d8ce	tweak	2022-10-31 18:19:43 +00:00
Starbeamrainbowlabs	dbe35ee943	loss: comment l2 norm	2022-10-31 18:09:03 +00:00
Starbeamrainbowlabs	5e60319024	fixup	2022-10-31 17:56:49 +00:00
Starbeamrainbowlabs	b986b069e2	debug party time	2022-10-31 17:50:29 +00:00
Starbeamrainbowlabs	458faa96d2	loss: fixup	2022-10-31 17:18:21 +00:00
Starbeamrainbowlabs	55dc05e8ce	contrastive: comment weights that aren't needed	2022-10-31 16:26:48 +00:00
Starbeamrainbowlabs	48ae8a5c20	LossContrastive: normalise features as per the paper	2022-10-26 16:52:56 +01:00
Starbeamrainbowlabs	843cc8dc7b	contrastive: rewrite the loss function. ... The CLIP paper *does* kinda make sense I think	2022-10-26 16:45:45 +01:00
Starbeamrainbowlabs	fad1399c2d	convnext: whitespace	2022-10-26 16:45:20 +01:00
Starbeamrainbowlabs	1d872cb962	contrastive: fix initial temperature value ... It should be 1/0.07, but we had it set to 0.07......	2022-10-26 16:45:01 +01:00
Starbeamrainbowlabs	f994d449f1	Layer2Image: fix	2022-10-25 21:32:17 +01:00
Starbeamrainbowlabs	6a29105f56	model_segmentation: stack not reshape	2022-10-25 21:25:15 +01:00
Starbeamrainbowlabs	98417a3e06	prepare for NCE loss ... .....but Tensorflow's implementation looks to be for supervised models :-(	2022-10-25 21:15:05 +01:00
Starbeamrainbowlabs	f2e2ca1484	model_contrastive: make water encoder significantly shallower	2022-10-24 20:52:31 +01:00
Starbeamrainbowlabs	6423bf6702	LayerConvNeXtGamma: avoid adding an EagerTensor to config ... Very weird how this is a problem when it wasn't before..	2022-10-12 17:12:07 +01:00
Starbeamrainbowlabs	c45b90764e	segmentation: adds xxtiny, but unsure if it's small enough	2022-10-11 19:22:37 +01:00
Starbeamrainbowlabs	11f91a7cf4	train: add --arch; default to convnext_i_xtiny	2022-10-11 19:18:01 +01:00
Starbeamrainbowlabs	2960d3b645	exception → warning	2022-10-06 18:26:40 +01:00
Starbeamrainbowlabs	0ee6703c1e	Add todo and comment	2022-10-03 19:06:56 +01:00
Starbeamrainbowlabs	2b182214ea	typo	2022-10-03 17:53:10 +01:00
Starbeamrainbowlabs	92c380bff5	fiddle with Conv2DTranspose ... you need to set the `stride` argument to actually get it to upscale..... :P	2022-10-03 17:51:41 +01:00
Starbeamrainbowlabs	fc6d2dabc9	Upscale first, THEN convnext...	2022-10-03 16:38:43 +01:00
Starbeamrainbowlabs	6a0790ff50	convnext_inverse: add returns; change ordering	2022-10-03 16:32:09 +01:00
Starbeamrainbowlabs	7130c4fdf8	start implementing core image segmentation model	2022-09-07 17:45:38 +01:00
Starbeamrainbowlabs	22620a1854	ai: implement saving only the rainfall encoder	2022-09-06 19:48:46 +01:00
Starbeamrainbowlabs	3e13ad12c8	ai Bugfix LayerContrastiveEncoder: channels → input_channels ... for consistency	2022-09-05 23:53:16 +01:00