Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Run_svi tweaks #10

Open
wants to merge 1 commit into
base: dev
Choose a base branch
from
Open

Run_svi tweaks #10

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
50 changes: 43 additions & 7 deletions examples/run_svi.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,25 +10,40 @@
parser = argparse.ArgumentParser(description="Train DeepSequence with SVI.")
parser.add_argument("--dataset", type=str, default="BLAT_ECOLX",
help="Dataset name for fitting model.")
parser.add_argument("--neff-override", type=float, default=None,
help="Override the model Neff.")
parser.add_argument("--theta-override", type=float, default=None,
help="Override the model theta.")
parser.add_argument("--n-latent-override", type=int, default=None,
help="Override the model n_latent.")
parser.add_argument("--weights_dir", type=str, default="", help="Location of precomputed weights, if possible")
parser.add_argument("--alignments_dir", type=str, help="Location of alignments")
parser.add_argument("--seed", type=int, help="Random seed override (for model ensembling).")
args = parser.parse_args()

args.dataset = args.dataset.split(".a2m")[0]

data_params = {
"dataset" : args.dataset,
"dataset" : args.dataset,
"weights_dir" : args.weights_dir,
}

if args.seed is not None:
print("Using seed: {}".format(args.seed))

model_params = {
"bs" : 100,
"encode_dim_zero" : 1500,
"encode_dim_one" : 1500,
"decode_dim_zero" : 100,
"decode_dim_one" : 500,
"decode_dim_one" : 2000, # 500 in the repo
"n_latent" : 30,
"logit_p" : 0.001,
"sparsity" : "logit",
"final_decode_nonlin": "sigmoid",
"final_pwm_scale" : True,
"n_pat" : 4,
"r_seed" : 12345,
"r_seed" : args.seed if args.seed is not None else 12345,
"conv_pat" : True,
"d_c_size" : 40
}
Expand All @@ -37,13 +52,24 @@
"num_updates" : 300000,
"save_progress" : True,
"verbose" : True,
"save_parameters" : False,
"save_parameters" : 50000,
}

if __name__ == "__main__":
if args.n_latent_override:
model_params['n_latent'] = args.n_latent_override

if __name__ == "__main__":
start_time = time.time()
data_helper = helper.DataHelper(dataset=data_params["dataset"],
calc_weights=True)
working_dir='.',
calc_weights=False, # Use precomputed weights
theta=args.theta_override,
weights_dir=data_params["weights_dir"],
alignments_dir=args.alignments_dir,
)
print("Data loaded.")
if args.neff_override:
data_helper.Neff = args.neff_override

vae_model = model.VariationalAutoencoder(data_helper,
batch_size = model_params["bs"],
Expand All @@ -63,16 +89,26 @@
n_patterns = model_params["n_pat"],
random_seed = model_params["r_seed"],
)
print("Model loaded")

job_string = helper.gen_job_string(data_params, model_params)
if args.neff_override:
job_string += "_neff-" + str(args.neff_override)
if args.seed is not None:
job_string += "_seed-" + str(args.seed)

print ("job string: ", job_string)

print (job_string)
print("Starting training")

train.train(data_helper, vae_model,
num_updates = train_params["num_updates"],
save_progress = train_params["save_progress"],
save_parameters = train_params["save_parameters"],
verbose = train_params["verbose"],
job_string = job_string)
print("Training complete")

vae_model.save_parameters(file_prefix=job_string)

print("Done in " + str(time.time() - start_time) + " seconds")