[Feature]: add swanlab logger (#7)

.gitignore

@@ -7,4 +7,4 @@ datasets/
 experiments_t2m/
 experiments_t2m_test/
 experiments_control/
-experiments_control_test/
+experiments_control_test/

README.md

@@ -41,6 +41,7 @@ Experimental results demonstrate the remarkable generation and controlling capab
 
 ## 📢 News
 
+- **[2024/05/02]** We support the [SwanLab](https://github.com/SwanHubX/SwanLab) logger, please refer to this [PR](https://github.com/Dai-Wenxun/MotionLCM/pull/7) for details.
 - **[2024/05/01]** Upload paper and release code.
 
 ## 👨‍🏫 Quick Start

mld/config.py

@@ -35,6 +35,7 @@ def parse_args() -> DictConfig:
     parser.add_argument('--example', type=str, required=False, help="input text and lengths with txt format")
     parser.add_argument('--no-plot', action="store_true", required=False, help="whether plot the skeleton-based motion")
     parser.add_argument('--replication', type=int, default=1, help="the number of replication of sampling")
+    parser.add_argument('--vis', type=str, default="tb", choices=['tb', 'swanlab'], help="the visualization method, tensorboard or swanlab")
     args = parser.parse_args()
 
     cfg = OmegaConf.load(args.cfg)
@@ -44,4 +45,5 @@ def parse_args() -> DictConfig:
     cfg.example = args.example
     cfg.no_plot = args.no_plot
     cfg.replication = args.replication
+    cfg.vis = args.vis
     return cfg

requirements.txt

@@ -3,6 +3,7 @@ torch==1.13.1
 gdown
 omegaconf
 rich
+swanlab==0.3.1
 torchmetrics==1.3.2
 scipy==1.11.2
 matplotlib==3.3.4
@@ -14,4 +15,4 @@ h5py==3.11.0
 smplx==0.1.28
 chumpy==0.70
 numpy==1.23.1
-natsort==8.4.0
+natsort==8.4.0

train_motion_control.py

@@ -8,6 +8,7 @@
 from omegaconf import OmegaConf
 
 import torch
+import swanlab
 import diffusers
 import transformers
 from torch.utils.tensorboard import SummaryWriter
@@ -30,8 +31,14 @@ def main():
     output_dir = osp.join(cfg.FOLDER, name_time_str)
     os.makedirs(output_dir, exist_ok=False)
     os.makedirs(f"{output_dir}/checkpoints", exist_ok=False)
-
-    writer = SummaryWriter(output_dir)
+
+    if cfg.vis == "tb":
+        writer = SummaryWriter(output_dir)
+    elif cfg.vis == "swanlab":
+        run = swanlab.init(project="MotionLCM", experiment_name=os.path.normpath(output_dir).replace(os.path.sep, "-"),
+                           suffix=None, config=cfg, logdir=output_dir)
+    else:
+        raise ValueError(f"Invalid vis method: {cfg.vis}")
 
     stream_handler = logging.StreamHandler(sys.stdout)
     file_handler = logging.FileHandler(osp.join(output_dir, 'output.log'))
@@ -133,7 +140,10 @@ def validation():
         min_val_tj = metrics['Metrics/traj_fail_50cm']
         print_table(f'Metrics@Step-{global_step}', metrics)
         for k, v in metrics.items():
-            writer.add_scalar(k, v, global_step=global_step)
+            if cfg.vis == "tb":
+                writer.add_scalar(k, v, global_step=global_step)         
+            elif cfg.vis == "swanlab":            
+                run.log({k: v}, step=global_step)
 
         model.controlnet.train()
         model.traj_encoder.train()
@@ -189,7 +199,10 @@ def validation():
                     "diff_loss": diff_loss.detach().item(), 'cond_loss': cond_loss.detach().item(), 'rot_loss': rot_loss.detach().item()}
             progress_bar.set_postfix(**logs)
             for k, v in logs.items():
-                writer.add_scalar(k, v, global_step=global_step)
+                if cfg.vis == "tb":
+                    writer.add_scalar(k, v, global_step=global_step)         
+                elif cfg.vis == "swanlab":            
+                    run.log({k: v}, step=global_step)
 
             if global_step >= cfg.TRAIN.max_train_steps:
                 break

train_motionlcm.py

@@ -10,6 +10,7 @@
 from omegaconf import OmegaConf
 
 import torch
+import swanlab
 import diffusers
 import transformers
 import torch.nn.functional as F
@@ -127,7 +128,13 @@ def main():
     os.makedirs(output_dir, exist_ok=False)
     os.makedirs(f"{output_dir}/checkpoints", exist_ok=False)
 
-    writer = SummaryWriter(output_dir)
+    if cfg.vis == "tb":
+        writer = SummaryWriter(output_dir)
+    elif cfg.vis == "swanlab":
+        run = swanlab.init(project="MotionLCM", experiment_name=os.path.normpath(output_dir).replace(os.path.sep, "-"),
+                           suffix=None, config=cfg, logdir=output_dir)
+    else:
+        raise ValueError(f"Invalid vis method: {cfg.vis}")
 
     stream_handler = logging.StreamHandler(sys.stdout)
     file_handler = logging.FileHandler(osp.join(output_dir, 'output.log'))
@@ -245,7 +252,10 @@ def validation():
         min_val_fid = metrics['Metrics/FID']
         print_table(f'Metrics@Step-{global_step}', metrics)
         for k, v in metrics.items():
-            writer.add_scalar(k, v, global_step=global_step)
+            if cfg.vis == "tb":
+                writer.add_scalar(k, v, global_step=global_step)
+            elif cfg.vis == "swanlab":
+                run.log({k: v}, step=global_step)
         base_model.train()
         return max_val_rp1, min_val_fid
 
@@ -411,9 +421,12 @@ def validation():
 
             logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
             progress_bar.set_postfix(**logs)
-            writer.add_scalar('loss', logs['loss'], global_step=global_step)
-            writer.add_scalar('lr', logs['lr'], global_step=global_step)
-
+            if cfg.vis == "tb":
+                writer.add_scalar('loss', logs['loss'], global_step=global_step)
+                writer.add_scalar('lr', logs['lr'], global_step=global_step)
+            elif cfg.vis == "swanlab":
+                run.log({'loss': logs['loss'], 'lr': logs['lr']}, step=global_step)
+
             if global_step >= cfg.TRAIN.max_train_steps:
                 break