Update README.md

README.md

@@ -3,3 +3,16 @@
 [DDSP_Filters.py] does not depend on anything else here, and can be used by itself.
 
 The other files are tied to [experiment.py], which is setup to train a DDSP model with (hopefully) very little adjustment. It does require the LJSpeech dataset though (https://keithito.com/LJ-Speech-Dataset/).
+
+An example of how to run the experiment script:
+
+`python experiment.py --experiment vocoder_run_1 --hidden-dim 512 --n-harmonics 67 --n-filters 101 --gpus 1 --max_epochs 500 -bs 8 -lr 1e-4 --loss msmsl --noise-level 0.1 --upsample 8.5 --force_f0 --train_files 'filelists/ljs_audio_text_train_filelist.txt' --val_files 'filelists/ljs_audio_text_val_filelist.txt' --num_workers 4 --profiler simple --precision 16 --deterministic True  --gradient_clip_val 0.5 --progress_bar_refresh_rate 1 --log_every_n_steps 50 --flush_logs_every_n_steps 50`
+
+The following arguments also must be specified:
+
+`-o [output path]`
+
+`-d [location of LJSpeech]`
+
+`--default_root_dir [path for Tensorboard logs]`
+