Show Menu
Cheatography

nvidia tlt Cheat Sheet by

for nvidia transfer learning

Introd­uction

NVIDIA Transfer Learning Toolkit (TLT) is a simple, easy-t­o-use training toolkit that requires minimal to zero coding to create vision AI models using the user’s own data. This cheatsheet is created by Ness, version 2020.1­1.14.

tlt stack

tlt modles

tlt workflow

 

tlt-au­gment

-d /path/­to/­the­/da­tas­et/root
-a /path/­to/­aug­men­tat­ion­/sp­ec/file
-o /path/­to/­the­/au­gme­nte­d/o­utput
[-v]

tlt-da­tas­et-­convert

-d DATASE­T_E­XPO­RT_SPEC
-o OUTPUT­_FI­LENAME
[-f VALIDA­TIO­N_FOLD]

spec files

 

tlt-train

classi­fic­ation --gpus <num GPUs>
-k <en­coding key>
-r <result direct­ory>
-e <spec file>

tlt-ev­aluate classi­fic­ation

-e <ex­per­ime­nt_­spe­c_f­ile>
-k <ke­y>

tlt-ev­aluate detect­net_v2

-e <ex­per­ime­nt_­spe­c_f­ile>
-m <mo­del­_fi­le>
-k <ke­y>
[--use­_tr­ain­ing­_set]

tlt-prune

tlt-prune [-h]
-pm <pr­etr­ain­ed_­mod­el>
-o <ou­tpu­t_f­ile> -k <ke­y>
[-n <no­rma­liz­er>]
[-eq <eq­ual­iza­tio­n_c­rit­eri­on>]
[-pg <pr­uni­ng_­gra­nul­ari­ty>]
[-pth <pr­uning thresh­old­>]
[-nf <mi­n_n­um_­fil­ter­s>]
[-el [<e­xcl­ude­d_l­ist­>]
 

tlt-in­t8-­ten­sorfile

tlt-in­t8-­ten­sorfile {class­ifi­cation, detect­net_v2} [-h]
-e <path to training experiment spec file>
-o <path to output tensor­fil­e>
-m <ma­ximum number of batches to serial­ize>
[--use­_va­lid­ati­on_set]

tlt-export

tlt-export [-h] {class­ifi­cation, detect­net_v2, ssd, dssd, faster­_rcnn, yolo, retinanet}
-m <path to the .tlt model file generated by tlt train>
-k <ke­y>
[-o <path to output file>]
[--cal­_da­ta_file <path to tensor file>]
[--cal­_im­age_dir <path to the directory images to calibrate the model]
[--cal­_ca­che­_file <path to output calibr­ation file>]
[--dat­a_type <Data type for the TensorRT backend during export­>]
[--batches <Number of batches to calibrate over>]
[--max­_ba­tch­_size <ma­ximum trt batch size>]
[--max­_wo­rks­pac­e_size <ma­ximum workspace size]
[--bat­ch_size <batch size to TensorRT engine­>]
[--exp­eri­men­t_spec <path to experiment spec file>]
[--eng­ine­_file <path to the TensorRT engine file>]
[--verbose Verbosity of the logger]
[--for­ce_ptq Flag to force PTQ]
 

Comments

No comments yet. Add yours below!

Add a Comment

Your Comment

Please enter your name.

    Please enter your email address

      Please enter your Comment.