File size: 6,359 Bytes
83034b6 |
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 |
import wandb
import os
import matplotlib.pyplot as plt
import random
is_wandb_enabled = os.getenv("WANDB_DISABLED", "false").lower() != "true"
def get_next_directory(base_dir="results", sub_dir="train"):
result_dir = os.path.join(base_dir, sub_dir)
if not os.path.exists(result_dir):
os.makedirs(result_dir)
run_id = 0
while True:
run_dir = os.path.join(result_dir, f"VAL_Images_StyleTransfer_{run_id}")
if not os.path.exists(run_dir):
os.makedirs(run_dir)
return run_dir, run_id
run_id += 1
def init_project(base_dir="results", sub_dir="train"):
project_dir, run_id = get_next_directory(base_dir, sub_dir)
if is_wandb_enabled:
# `name` là cố định, còn `project` sẽ là VAL_Images_StyleTransfer_{run_id}
wandb.init(project=f"VAL_Images_StyleTransfer", name=f"VAL_Images_StyleTransfer_{run_id}r")
return project_dir
def upload_images(images, iters, img_index, epoch):
dir_name = f"epoch_{epoch}_iters_{iters}"
if not os.path.exists(dir_name):
os.makedirs(dir_name)
for i, img in enumerate(images):
img_name = f"img_{img_index}_{iters}_{epoch}.png"
img_path = os.path.join(dir_name, img_name)
plt.figure()
plt.imshow(img)
plt.title(f"Image {img_index} | Iter {iters} | Epoch {epoch}")
plt.axis('off')
plt.savefig(img_path)
plt.close()
if is_wandb_enabled:
wandb.log({f"epoch_{epoch}/img_{img_index}_{iters}_{epoch}": wandb.Image(img_path)})
#### ON ONE IMAGES
previous_iters_all_images = -1
last_iters=0
def reset_iters_if_needed(iters_all_images):
global previous_iters_all_images
if iters_all_images > previous_iters_all_images:
previous_iters_all_images = iters_all_images
return True # Cần reset cho section mới
return False
def get_section_name(epoch, iters_all_images):
return f"Iters on all Images_epoch{epoch}_iters{iters_all_images}"
def define_wandb_metrics(epoch, iters_all_images):
section_name = get_section_name(epoch, iters_all_images)
if is_wandb_enabled:
# Định nghĩa metric cho từng section (epoch + iters_all_images)
wandb.define_metric(f"{section_name}/All_loss_on_one_images", step_metric=f"{section_name}_step")
wandb.define_metric(f"{section_name}/L1_loss_on_one_images", step_metric=f"{section_name}_step")
wandb.define_metric(f"{section_name}/L2_loss_on_one_images", step_metric=f"{section_name}_step")
wandb.define_metric(f"{section_name}/Content_loss_on_one_images", step_metric=f"{section_name}_step")
wandb.define_metric(f"{section_name}/Style_loss_on_one_images", step_metric=f"{section_name}_step")
def upload_all_loss_on_one_images(loss_value, iters_one_imgs, iters_all_images, epoch):
if reset_iters_if_needed(iters_all_images):
iters_one_imgs = last_iters+iters_one_imgs
define_wandb_metrics(epoch, iters_all_images) # Định nghĩa các metric cho section mới
section_name = get_section_name(epoch, iters_all_images)
if is_wandb_enabled:
wandb.log({f"{section_name}/All_loss_on_one_images": loss_value, f"{section_name}_step": iters_one_imgs})
def upload_l1_loss_on_one_images(loss_value, iters_one_imgs, iters_all_images, epoch):
section_name = get_section_name(epoch, iters_all_images)
if is_wandb_enabled:
wandb.log({f"{section_name}/L1_loss_on_one_images": loss_value, f"{section_name}_step": iters_one_imgs})
def upload_l2_loss_on_one_images(loss_value, iters_one_imgs, iters_all_images, epoch):
section_name = get_section_name(epoch, iters_all_images)
if is_wandb_enabled:
wandb.log({f"{section_name}/L2_loss_on_one_images": loss_value, f"{section_name}_step": iters_one_imgs})
def upload_content_loss_on_one_images(loss_value, iters_one_imgs, iters_all_images, epoch):
section_name = get_section_name(epoch, iters_all_images)
if is_wandb_enabled:
wandb.log({f"{section_name}/Content_loss_on_one_images": loss_value, f"{section_name}_step": iters_one_imgs})
def upload_style_loss_on_one_images(loss_value, iters_one_imgs, iters_all_images, epoch):
section_name = get_section_name(epoch, iters_all_images)
if is_wandb_enabled:
wandb.log({f"{section_name}/Style_loss_on_one_images": loss_value, f"{section_name}_step": iters_one_imgs})
### ON ALL IMAGES
def upload_l1_loss_all(loss_values, iters_images, epoch):
if is_wandb_enabled:
wandb.log({f"epoch_{epoch}/L1_loss": loss_values}, step=iters_images)
def upload_l2_loss(loss_values, iters, epoch):
if is_wandb_enabled:
wandb.log({f"epoch_{epoch}/L2_loss": loss_values}, step=iters)
def upload_content_loss(loss_c, iters, epoch):
if is_wandb_enabled:
wandb.log({f"epoch_{epoch}/Content_loss": loss_c}, step=iters)
def upload_style_loss(loss_s, iters, epoch):
if is_wandb_enabled:
wandb.log({f"epoch_{epoch}/Style_loss": loss_s}, step=iters)
def upload_epoch(epoch,iter_images):
if is_wandb_enabled:
wandb.log({"epoch": epoch})
def upload_lr(lr,epoch):
if is_wandb_enabled:
wandb.log({"learning_rate": lr})
def upload_loss_all(loss_all, iters, epoch):
if is_wandb_enabled:
wandb.log({f"epoch_{epoch}/Total_loss": loss_all}, step=iters)
def logs_model(model_params):
"""Ghi log các tham số của mô hình lên Wandb."""
if is_wandb_enabled:
wandb.log({"model_parameters": model_params})
def logs_params(params):
"""Ghi log các tham số khác lên Wandb."""
if is_wandb_enabled:
wandb.log({"params": params})
# def save_model_weights(model, epoch, project_dir):
# """Lưu weights của model theo từng epoch."""
# weights_dir = os.path.join(project_dir, f"epoch_{epoch}")
# if not os.path.exists(weights_dir):
# os.makedirs(weights_dir)
#
# weights_path = os.path.join(weights_dir, f"weights_epoch_{epoch}.pth")
# torch.save(model.state_dict(), weights_path) # Lưu weights vào file .pth
#
# if is_wandb_enabled:
# wandb.save(weights_path)
def logs_running(terminal_output):
if is_wandb_enabled:
wandb.log({"terminal_output": wandb.Html(terminal_output)})
def stop():
if is_wandb_enabled:
wandb.finish()
|