me/ml-audio-enhancer
Watch
1
0
Fork
You've already forked ml-audio-enhancer
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Update README.md

Browse source
This commit is contained in:
Andrew Gundersen 2020-02-17 17:37:41 -06:00 • committed by GitHub
commit 7cfb8e071d
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 23 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

44
README.md
View file

@ -5,6 +5,10 @@ To read more about the research, visit https://crimata.com/AIPost
## Usage ## Usage
Clone the repo.
Download dataset at https://homepages.inf.ed.ac.uk/jyamagis/page3/page58/page58.html and place it in the data folder.
Repository Structure Repository Structure
* `./data`: the data and the metadata (you should not need to touch this folder) * `./data`: the data and the metadata (you should not need to touch this folder)
@ -26,31 +30,29 @@ python 3 main.py eval -h
Argument Reference | train Argument Reference | train
``` ```
Arguments: -h, --help
-h, --help -i MODEL_ID an index you can add to querey the model later
-i MODEL_ID an index you can add to querey the model later -c FROM_CKPT bool, begin training session from a checkpoint
-c FROM_CKPT bool, begin training session from a checkpoint -k NEW_DATA bool, make new set of data
-k NEW_DATA bool, make new set of data -d DIM_SIZE size of each audio sample
-d DIM_SIZE size of each audio sample -x NUM_FILES number of files to train
-x NUM_FILES number of files to train -e EPOCHS
-e EPOCHS -b BATCH_SIZE
-b BATCH_SIZE -o CYCLE_LENGTH length of cycle in epochs (only for SGDR)
-o CYCLE_LENGTH length of cycle in epochs (only for SGDR) -m MAX_LR max learning rate (only for SGDR)
-m MAX_LR max learning rate (only for SGDR) -n MIN_LR min learning rate (only for SGDR)
-n MIN_LR min learning rate (only for SGDR)
``` ```
Argument Reference | eval Argument Reference | eval
``` ```
Arguments: -h, --help
-h, --help -i MODEL_ID model ID to use
-i MODEL_ID model ID to use -n NUM_EXAMPLES number of examples/files to run
-n NUM_EXAMPLES number of examples/files to run -w WAVFILE_LIST list to pull examples from
-w WAVFILE_LIST list to pull examples from -r SCALE level of upsampling
-r SCALE level of upsampling -s SAMPLE_RATE target sample rate
-s SAMPLE_RATE target sample rate -a MAKE_AUDIO bool, make audio or just spectrograms
-a MAKE_AUDIO bool, make audio or just spectrograms -c FROM_CKPT bool, begin eval at last checkpoint
-c FROM_CKPT bool, begin eval at last checkpoint
``` ```