-
Notifications
You must be signed in to change notification settings - Fork 0
/
my_main_test.py
194 lines (150 loc) · 7.62 KB
/
my_main_test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
import numpy as np
import pandas as pd
import os
import time
import numpy
import torch
import torch.optim as optim
import yaml
from torch.utils.data import DataLoader
from utils import criterion_utils, data_utils, eval_utils, model_utils
from eval_script import retrieval_metrics
# from models import ema
import h5py
torch_seed=0
numpy.random.seed(42)
torch.manual_seed(torch_seed)
# YAML_FPATH = "choro_conf.yaml"
YAML_FPATH = "my_conf.yaml"
with open(YAML_FPATH, "rb") as stream:
config = yaml.full_load(stream)
# model ckpt directory
# checkpoint_dir='best_checkpoints/checkpoint_TAGS_sigmoidAS_0.800_margin_0.4_000026_mAP_2342'
# checkpoint_dir='best_checkpoints/checkpoint_TAGS_sigmoidAS_0.800_margin_0.4_000018_mAP_2339'
# checkpoint_dir='best_checkpoints/checkpoint_seed2_TAGS_sigmoidAS_0.800_margin_0.4_000021_mAP_2323'
# checkpoint_dir='best_checkpoints/checkpoint_pretrainedAudioCaps_seed1978_seed0_TAGS_sigmoidAS_0.800_margin_0.4_000053_mAP_2396'
# checkpoint_dir='best_checkpoints/checkpoint_pretrainedAudioCaps_seed1978_seed42_TAGS_sigmoidAS_0.800_margin_0.4_noScheduler_000043_mAP_2402'
checkpoint_dir='outputs_passt/checkpoint_margin1_20juin2022/checkpoint_000030'
# checkpoint_dir='best_checkpoints/checkpoint_seed0_none_margin_0.4_000018_mAP_2292'
# checkpoint_dir='best_checkpoints/checkpoint_TAGS_sigmoidAS_0.800_margin_0.4_000018_mAP_2307'
# checkpoint_dir='best_checkpoints/checkpoint_TAGS_sigmoidAS_0.800_margin_0.4_000021'
training_config = config["training"]
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
use_sentence_embeddings = True
use_auto_caption_embeddings = False
print("Using Etienne's auto caption embeddings?", use_auto_caption_embeddings)
# load Clotho data
# text_datasets, vocabulary = data_utils.load_data(config["train_data"])
caption_datasets, vocabulary = data_utils.load_data(config["eval_data"])
model_config = config[training_config["model"]]
model = model_utils.get_model(model_config, vocabulary)
print(model)
# print(model.audio_encoder.pointwise_conv.weight.data)
# model_state, optimizer_state = torch.load(os.path.join(checkpoint_dir, "checkpoint"))
# model.load_state_dict(model_state)
# optimizer.load_state_dict(optimizer_state)
# for state in optimizer.state.values():
# for k, v in state.items():
# if isinstance(v, torch.Tensor):
# state[k] = v.to(device)
#
# Restore model states
model = model_utils.restore(model, checkpoint_dir)
model.eval()
print("Nb of learnable parameters :",
sum(p.numel() for p in model.parameters() if p.requires_grad))
# alg_config = training_config["algorithm"]
#
# criterion_config = config[alg_config["criterion"]]
# if config["train_data"]["use_mixup"] and criterion_config["name"] != "MixupTripletRankingLoss":
# print("ERROR: change criterion to MixupTripletRankingLoss in config file")
#
#
# criterion = getattr(criterion_utils, criterion_config["name"], None)(**criterion_config["args"])
#
# optimizer_config = config[alg_config["optimizer"]]
# optimizer = getattr(optim, optimizer_config["name"], None)(
# model.parameters(), **optimizer_config["args"]
# )
# lr_scheduler = getattr(optim.lr_scheduler, "ReduceLROnPlateau")(optimizer, **optimizer_config["scheduler_args"])
clotho_split = "test"
clotho_dataset = caption_datasets[clotho_split]
# Dataloader not needed since files are processed one by one
# clotho_loader = DataLoader(dataset=clotho_dataset, batch_size=training_config["algorithm"]["batch_size"],
# shuffle=False, collate_fn=data_utils.collate_fn_sentence_embeddings)
# Retrieve audio files for evaluation captions
# output = eval_utils.audio_retrieval(model, caption_datasets[clotho_split], K=10,
# use_sentence_embeddings=use_sentence_embeddings,
# use_auto_caption_embeddings=use_auto_caption_embeddings)
scores_h5_output_file = os.path.join(checkpoint_dir, "{}_scores.h5".format(clotho_split))
# scores_h5_output_file = None
output = eval_utils.audio_retrieval(model, clotho_dataset, K=10,
use_sentence_embeddings=use_sentence_embeddings,
use_auto_caption_embeddings=use_auto_caption_embeddings,
scores_output_fpath=scores_h5_output_file)
csv_fields = ["caption",
"file_name_1",
"file_name_2",
"file_name_3",
"file_name_4",
"file_name_5",
"file_name_6",
"file_name_7",
"file_name_8",
"file_name_9",
"file_name_10"]
output = pd.DataFrame(data=output, columns=csv_fields)
output.to_csv(os.path.join(checkpoint_dir, "{}.output.csv".format(clotho_split)),
index=False)
print("Saved", "{}.output.csv".format(clotho_split))
gt_csv = "test.gt.csv" # ground truth for Clotho evaluation data
pred_csv = os.path.join(checkpoint_dir,
"{}.output.csv".format(clotho_split)) # baseline system retrieved output for Clotho evaluation data
retrieval_metrics(gt_csv, pred_csv, log_wandb=False, is_ema=False)
# get loss values on train, val and test
text_datasets, vocabulary = data_utils.load_data(config["train_data"])
text_loaders = {}
for split in ["train", "val", "test"]:
_dataset = text_datasets[split]
if use_sentence_embeddings:
if use_auto_caption_embeddings:
_loader = DataLoader(dataset=_dataset, batch_size=training_config["algorithm"]["batch_size"],
shuffle=True, collate_fn=data_utils.collate_fn_sentence_and_auto_caption_embeddings)
else:
_loader = DataLoader(dataset=_dataset, batch_size=training_config["algorithm"]["batch_size"],
shuffle=True, collate_fn=data_utils.collate_fn_sentence_embeddings)
else:
_loader = DataLoader(dataset=_dataset, batch_size=training_config["algorithm"]["batch_size"],
shuffle=True, collate_fn=data_utils.collate_fn)
text_loaders[split] = _loader
alg_config = training_config["algorithm"]
criterion_config = config[alg_config["criterion"]]
if config["train_data"]["use_mixup"] and criterion_config["name"] != "MixupTripletRankingLoss":
print("ERROR: change criterion to MixupTripletRankingLoss in config file")
exit(-1)
criterion = getattr(criterion_utils, criterion_config["name"], None)(**criterion_config["args"])
split='train'
data_loader = text_loaders[split]
for batch_idx, data in enumerate(data_loader):
# Get the inputs; data is a list of tuples (audio_feats, audio_lens, queries, query_lens, infos)
audio_feats, audio_lens, queries, query_lens, infos, caption_embeds = data
audio_feats, caption_embeds = audio_feats.to(device), caption_embeds.to(device)
# print(audio_feats.size())
# Forward
audio_embeds, query_embeds, audio_embeds2 = model(audio_feats, caption_embeds)
loss, list_scores = criterion(audio_embeds, query_embeds, infos, is_train_subset=True)
import matplotlib.pyplot as plt
anchor_scores = list_scores[0]
A_imp_scores = list_scores[1]
Q_imp_scores = list_scores[2]
print("anchor=%.1f A_imp_scores=%.3f Q_imp_scores=%.3f\n"%(np.mean(anchor_scores), np.mean(A_imp_scores), np.mean(A_imp_scores)))
# test: anchor=0.4 A_imp_scores=0.005 Q_imp_scores=0.005
# train: anchor=0.4 A_imp_scores=0.015 Q_imp_scores=0.015
fig, ax = plt.subplots()
ax.hist(anchor_scores, bins=20, rwidth=0.9, alpha=0.2, label='anchor')
ax.set_xlabel("Scores")
# ax.set_ylabel("")
ax.hist(A_imp_scores, bins=20, rwidth=0.9, alpha=0.2, label='A imp')
ax.hist(Q_imp_scores, bins=20, rwidth=0.9, alpha=0.2, label='Q imp')
ax.legend()
plt.savefig("histogram_scores_%s.png"%(split))