The opposite of scarcity is not abundance: saturation inevitably damages, or perhaps it just corrodes or blunts the impassioned pavlov puppy, delivering a dent of tiny deliberate delirious wonder.
Yet technical momentum now propels and impels continuance: how can the incoherence be tamed? Set some new hyperparameters and let the wvnt train for 26 hours.
Over this weekend, I’ve churned out about 100,000 lines. Generating reckless amounts of incoherent poetry threatens more than perspective or contemplation, it totters sanity on the whim of a machine. Teeming bacteria, every epiphany a whiff of redundancy.
$ python train_2017_py3p5_150k_low6_sample4096_SAVE-STARTS_100k.py --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --data_dir=data/2017 Using default logdir: ./logdir/train/2017-06-01T08-38-45 _______________________________________________________________________________________________ dilations: 2048 filter_width: 2 residual_channels: 32 dilation_channels: 32 skip_channels: 8192 quantization_channels: 2048
(tf0p1-py3.5-wvnt) jhave@jhave-UbuntuScreencast 2017-06-02 11:08:14_2017-06-01T08-38-45
and faithful wound To fruit white, the dread One by one another, Image saved-- Ay of the visit. What pursued my heart to brink. Such the curse of hopes fraught memory;
tf0p1-py3.5-wvnt_jhave-Ubuntu_Screencast 2017-06-02 11:08:14_2017-06-01T08-38-45
Caught a new light bulb, All the heart is grown.
TXTs generated in SLOW MODE
There’s a way of calculating the matrices that taxes the strength of evn a magnificient GPU, making production crawl, and the computer difficult to use. Each of the following txt files (4444 letters in each) took about 40-60 minutes to generate on an Nvidia Maxwell TitanX using cuda8 on Ubuntu 16.4
Txts generated slow seem somehow thicker, as if issued from a more calibrated mentation, yet at the same time it’s math scat, glitch flow. Glisses from disintegrating encyclopedias.
Here are some samples:
I found myself within us wonder. You purchase as ease with water over events, because the straightforward that I miximally, she Don't sports commentation with its ruffled story
There is a woefu's rupped, pelleys glow whud not eat resemblance. Phen my friends talking me numb because the hairspana of "strange tables. Oh, edgy and buckeying me with her children strion. Every one outside also that Who reaped down a fast-puliforous as dirty factions.
2017-06-03_17-19_DIR-2017-06-01T08-38-45_Model-112640_Loss-0.584_Chars-4444.
CLUMPED TO A SNAIL, gold-stiffened and rear it's 48 decraried A wilderness is snow bathed around you'd be delicacy their rejected and 19-purg foot unjust I have time it dow to keep thin age, just fine Cooper--feel This more, & undressed)
2017-06-03_14-29_DIR-2017-06-01T08-38-45_Model-117221_Loss-0.493_Chars-4444
terrible set up toward the wave. Assume, the evening, my own real equal, flick peders roam bed and reasonings have notched; they solidated us to turn shrouded by the metal of gray as I thought so headpan dust. o mother: all the madhouse that day this is my head.
2017-06-03_15-33_DIR-2017-06-01T08-38-45_Model-112608_Loss-0.230_Chars-4444
in its swollen fresh and healed the window creating. Bright, gelwired I sut lone there yet I'd give home of you. So on the land. I hold over you're like a disk.
2017-06-03_16-23_DIR-2017-06-01T08-38-45_Model-112609_Loss-0.243_Chars-4444
and you make eyes girl jokes Tristrain spider, as the sides of the far side plump of a frozen floor blanked what reaching or as wholly stairs for the dew splatter pail, an inteclined cheek together made, & shibboring the skull petal double as this aloft has a daily and lists kearted on the pillow.
2017-06-03_18-19_DIR-2017-06-01T08-38-45_Model-112641_Loss-0.432_Chars-4444
AND HIGH WORD-- Who sets mes chadmories of legities that read sickness, ettachbust subtractor's cigarette at thoroughness, a blinding the valleys followed me and before. Through the Drugot-- if the versions moved onto a perilous toast an innime sovereign rivers were for years-- intricate him, onhered attitude in flints, in the thin-so-wh, we stood ran striped his mothers. It was then they escaped to the stars.Above the Brotherford The years I know this guy - Where everyone was close. If Jr. tumknat, I'm everywhere, though, shield did, redress of now. Our grandchild (the two goes spring deep. Remake the historic part of the sob won't nursed only the engine to her work, a shooting angel with ink.
2017-06-03_19-11_DIR-2017-06-01T08-38-45_Model-112642_Loss-0.550_Chars-4444
Dawn Day forgive A grisse with jacks on her handed sheep, not just war from every tongues, its stem-ton fields. But if I touched the sound against the screw hooks tossed the ghost of means its pink inside the desire of whom I would injaid for this silence.
2017-06-03_20-02_DIR-2017-06-01T08-38-45_Model-112643_Loss-0.521_Chars-4444
If, one night fall in everyone's arms, as now it cold, no sickness what I'll be you think you've admit that after you packed. This drought through the soul, the white people erect love up by summer again: tin a stillness, with the shore and squinting cloud from the other hands me
2017-06-03_21-13_DIR-2017-06-01T08-38-45_Model-112859_Loss-0.579_Chars-4444
the eposits. Off of Kansay wasn't waste. Strong round move stood in frostmates, Ray Raffae. Never on thinking of Middle. Rips, unmarred by accident neckbelled bathrooms & daughter, a reprievement. Wailed, and Richard World, she is gone, though question is shadows, divine. What does it even us with and going the rest of headlights brought to slowly upward face while I understand and iides wordling out of their knife.
2017-06-03_22-07_DIR-2017-06-01T08-38-45_Model-113546_Loss-0.514_Chars-4444
Those ordinary warfare flesh,
I saw me many weeping, until we stayed
the innocence of a clued out with)
dark flanks
2017-06-04_07-36_DIR-2017-06-01T08-38-45_Model-122032_Loss-0.242_Chars-4444
The boats were sewn into the night--
2017-06-04_08-27_DIR-2017-06-01T08-38-45_Model-122033_Loss-0.338_Chars-4444
You keep books & sex cancels--drawing else abandoned. I want it beside a furnace, ineans and a guide of slow of enduring
And then a txt file containing 555-letter samples from all the models made on 06-01T08-38-45:
(tf0p1-py3.5-wvnt) jhave@jhave-Ubuntu:~2017-06-04_09-21-18-555chars-SLOW
was being his trances of mances of those As the furrow, hard-self father and thoughts, became necks at a tiny joy, a lark of iron on my bare stone. A shooting shadows came for joy; a hammer impossible body for them on town odor eclipse of scour lips the ocean loose, pulled down so long today is tiny benches and white anger, flat underground house, symbolist.
$ bash Run_Model_Demos_100K+-INFINITE_2017-GOOD-06-01T08-38-45-passSamples.sh 555 python -W ignore generate_Poems_2017-wsuppressed-PERPETUAL-SLOW.py --samples=4444 --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --loss=0.493 ./logdir/train/2017-06-01T08-38-45/model.ckpt-117221 python -W ignore generate_Poems_2017-wsuppressed-PERPETUAL-SLOW.py --samples=4444 --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --loss=0.230 ./logdir/train/2017-06-01T08-38-45/model.ckpt-112608 python -W ignore generate_Poems_2017-wsuppressed-PERPETUAL-SLOW.py --samples=4444 --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --loss=0.243 ./logdir/train/2017-06-01T08-38-45/model.ckpt-112609 python -W ignore generate_Poems_2017-wsuppressed-PERPETUAL-SLOW.py --samples=4444 --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --loss=0.584 ./logdir/train/2017-06-01T08-38-45/model.ckpt-112640 python -W ignore generate_Poems_2017-wsuppressed-PERPETUAL-SLOW.py --samples=4444 --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --loss=0.550 ./logdir/train/2017-06-01T08-38-45/model.ckpt-112642 python -W ignore generate_Poems_2017-wsuppressed-PERPETUAL-SLOW.py --samples=4444 --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --loss=0.242 ./logdir/train/2017-06-01T08-38-45/model.ckpt-122032 python -W ignore generate_Poems_2017-wsuppressed-PERPETUAL-SLOW.py --samples=4444 --wavenet_params=wavenet_params_ORIG_dilations2048_skipChannels8192_qc2048_dc32.json --loss=0.338 ./logdir/train/2017-06-01T08-38-45/model.ckpt-122033