python - Multiple issues with axes while implementing a Seq2Seq with attention in CNTK -


i'm trying implement seq2seq model attention in cntk, similar cntk tutorial 204. however, several small differences lead various issues , error messages, don't understand. there many questions here, interconnected , stem single thing don't understand.

note (in case it's important). input data comes minibatchsourcefromdata, created numpy arrays fit in ram, don't store in ctf.

ins = c.sequence.input_variable(input_dim, name="in", sequence_axis=inaxis) y = c.sequence.input_variable(label_dim, name="y", sequence_axis=outaxis)

thus, shapes [#, *](input_dim) , [#, *](label_dim).

question 1: when run cntk 204 tutorial , dump graph .dot file using cntk.logging.plot, see input shapes [#](-2,). how possible?

  • where did sequence axis (*) disappear?
  • how can dimension negative?

question 2: in same tutorial, have attention_axis = -3. don't understand this. in model there 2 dynamic axis , 1 static, "third last" axis #, batch axis. attention shouldn't computed on batch axis.
hoped looking @ actual axes in tutorial code me understand this, [#](-2,) issue above made more confusing.

setting attention_axis -2 gives following error:

runtimeerror: times: left operand 'placeholder('stab_result', [#, outaxis], [128])'               rank (1) must >= #axes (2) being reduced over.

during creation of training-time model: 

def train_model(m):     @c.function     def model(ins: inputsequence[tensor[input_dim]],                                 labels: outputsequence[tensor[label_dim]]):         past_labels = delay(initial_state=c.constant(seq_start_encoding))(labels)         return m(ins, past_labels)  #<<<<<<<<<<<<<< here     return model

where stab_result stabilizer right before final dense layer in decoder. can see in dot-file there spurious trailing dimensions of size 1 appear in middle of attentionmodel implementation.

setting attention_axis -1 gives following error:

runtimeerror: binary elementwise operation elementtimes: left operand 'output('block346442_output_0', [#, outaxis], [64])'               shape '[64]' not compatible right operand                'output('attention_weights', [#, outaxis], [200])' shape '[200]'.

where 64 attention_dim , 200 attention_span. understand, elementwise * inside attention model shouldn't conflating these 2 together, therefore -1 not right axis here.

question 3: understanding above correct? should right axis , why causing 1 of 2 exceptions above?

thanks explanations!

first, news: couple of things have been fixed in attentionmodel in latest master (will available cntk 2.2 in few days):

  • you don't need specify attention_span or attention_axis. if don't specify them , leave them @ default values, attention computed on whole sequence. in fact these arguments have been deprecated.
  • if above 204 notebook runs 2x faster, 204 notebook not use these arguments anymore
  • a bug has been fixed in attentionmodel , faithfully implements bahdanau et. al. paper.

regarding questions:

the dimension not negative. use negative numbers in various places mean things: -1 dimension inferred once based on first minibatch, -2 think shape of placeholder, , -3 dimension inferred each minibatch (such when feed variable sized images convolutions). think if print graph after first minibatch, should see shapes concrete.

attention_axis implementation detail should have been hidden. attention_axis=-3 create shape of (1, 1, 200), attention_axis=-4 create shape of (1, 1, 1, 200) , on. in general more -3 not guaranteed work , less -3 adds more 1s without clear benefit. news of course can ignore argument in latest master.

tl;dr: if in master (or starting cntk 2.2 in few days) replace attentionmodel(attention_dim, attention_span=200, attention_axis=-3) attentionmodel(attention_dim). faster , not contain confusing arguments. starting cntk 2.2 original api deprecated.


Comments

Post a Comment

Popular posts from this blog

angular - Ionic slides - dynamically add slides before and after -

hi i'm using ngfor create set of 3 slides while starting in middle i'm guaranteed able slide left or right on start. when slide right can simple listen reachedend , push slide array i'm looping. but have problem adding slide beginning. if same above , use e.g. array.unshift() or spread add item beginning, view think it's on position 0 , snaps view new slide. the code below work animates slide change index 1. slide = [0,1,2] //example loop slidechanged(event) { if(this.slides.isbeginning()){ this.slide = [this.slide[0]-1, ...this.slide]; this.slides.update(); this.slides.slideto(1) } } <ion-slides [initialslide]="1" (ionslidedidchange)="slidechanged($event)"> <ion-slide *ngfor="let item of slide"> <h1>slide {{item}}</h1> </ion-slide> </ion-slides> any appreciated! you can using ionslidenextend , ionslideprevend events slides . please
Read more

minify - Minimizing css files -

i want know how css files can minimized. currently, using external sites css minifier compress css files. however, after removing whitespaces, comments, etc, file sizes reduced 10-15%. how frameworks bootstrap manage make minimized versions kilobytes. as example, css is: body { background-color: #ff0; } the reduction goes 28 bytes 22 bytes other css files have lot more characters still smaller this. actually minifying code means it's remove thing code white space characters new line characters comments block delimiters it reduces amount of code has transferred on web , speed site load fastly.but unreadable format.
Read more

Add a dynamic header in angular 2 http provider -

i have http interceptor built via tutorial https://scotch.io/@kashyapmukkamala/using-http-interceptor-with-angular2 the problem need have dynamic header changes automatically once changes 'mode' in app. service able tell interceptor change header, or have interceptor pull header (which string)from service each request. able throwing away patch method, call http.patch('new header string') set header string. call in service controls 'mode' setting. wondering if there way interceptor pull recent value service each request or if can add method http provider via interceptor? i've tried call service in interceptor each request returns undefined. because creating new instance of service or copying value before service fired , loaded/changed data. i don't want pass header string every http provider call because defeats purpose of having interceptor, id able set once changes or have provider fetch service before makes request. tl;dr : need add autom
Read more