fix gradi_space

MuseV/scripts/gradio/gradio_text2video.py

@@ -69,7 +69,7 @@ args_dict = {
     "context_stride": 1,
     "cross_attention_dim": 768,
     "face_image_path": None,
-    "facein_model_cfg_path": "../../configs/model/facein.py",
+    "facein_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/facein.py"),
     "facein_model_name": None,
     "facein_scale": 1.0,
     "fix_condition_images": False,
@@ -82,14 +82,18 @@ args_dict = {
     "img_length_ratio": 1.0,
     "img_weight": 0.001,
     "interpolation_factor": 1,
-    "ip_adapter_face_model_cfg_path": "../../configs/model/ip_adapter.py",
+    "ip_adapter_face_model_cfg_path": os.path.join(
+        PROJECT_DIR, "configs/model/ip_adapter.py"
+    ),
     "ip_adapter_face_model_name": None,
     "ip_adapter_face_scale": 1.0,
-    "ip_adapter_model_cfg_path": "../../configs/model/ip_adapter.py",
+    "ip_adapter_model_cfg_path": os.path.join(
+        PROJECT_DIR, "configs/model/ip_adapter.py"
+    ),
     "ip_adapter_model_name": "musev_referencenet",
     "ip_adapter_scale": 1.0,
     "ipadapter_image_path": None,
-    "lcm_model_cfg_path": "../../configs/model/lcm_model.py",
+    "lcm_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/lcm_model.py"),
     "lcm_model_name": None,
     "log_level": "INFO",
     "motion_speed": 8.0,
@@ -101,7 +105,7 @@ args_dict = {
     "need_img_based_video_noise": True,
     "need_redraw": False,
     "negative_prompt": "V2",
-    "negprompt_cfg_path": "../../configs/model/negative_prompt.py",
+    "negprompt_cfg_path": os.path.join(PROJECT_DIR, "configs/model/negative_prompt.py"),
     "noise_type": "video_fusion",
     "num_inference_steps": 30,
     "output_dir": "./results/",
@@ -115,22 +119,26 @@ args_dict = {
     "redraw_condition_image_with_ipdapter": True,
     "redraw_condition_image_with_referencenet": True,
     "referencenet_image_path": None,
-    "referencenet_model_cfg_path": "../../configs/model/referencenet.py",
+    "referencenet_model_cfg_path": os.path.join(
+        PROJECT_DIR, "configs/model/referencenet.py"
+    ),
     "referencenet_model_name": "musev_referencenet",
     "save_filetype": "mp4",
     "save_images": False,
-    "sd_model_cfg_path": "../../configs/model/T2I_all_model.py",
+    "sd_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/T2I_all_model.py"),
     "sd_model_name": "majicmixRealv6Fp16",
     "seed": None,
     "strength": 0.8,
     "target_datas": "boy_dance2",
-    "test_data_path": "../../configs/infer/testcase_video_famous.yaml",
-    "time_size": 24,
-    "unet_model_cfg_path": "../../configs/model/motion_model.py",
+    "test_data_path": os.path.join(
+        PROJECT_DIR, "configs/infer/testcase_video_famous.yaml"
+    ),
+    "time_size": 12,
+    "unet_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/motion_model.py"),
     "unet_model_name": "musev_referencenet",
     "use_condition_image": True,
     "use_video_redraw": True,
-    "vae_model_path": "../../checkpoints/vae/sd-vae-ft-mse",
+    "vae_model_path": os.path.join(PROJECT_DIR, "checkpoints/vae/sd-vae-ft-mse"),
     "video_guidance_scale": 3.5,
     "video_guidance_scale_end": None,
     "video_guidance_scale_method": "linear",
@@ -138,7 +146,9 @@ args_dict = {
     "video_num_inference_steps": 10,
     "video_overlap": 1,
     "vision_clip_extractor_class_name": "ImageClipVisionFeatureExtractor",
-    "vision_clip_model_path": "../../checkpoints/IP-Adapter/models/image_encoder",
+    "vision_clip_model_path": os.path.join(
+        PROJECT_DIR, "checkpoints/IP-Adapter/models/image_encoder"
+    ),
     "w_ind_noise": 0.5,
     "width": None,
     "write_info": False,
@@ -248,17 +258,17 @@ n_repeat = args.n_repeat
 
 b = 1
 negative_embedding = [
-    ["../../checkpoints/embedding/badhandv4.pt", "badhandv4"],
+    [os.path.join(PROJECT_DIR, "checkpoints/embedding/badhandv4.pt"), "badhandv4"],
     [
-        "../../checkpoints/embedding/ng_deepnegative_v1_75t.pt",
+        os.path.join(PROJECT_DIR, "checkpoints/embedding/ng_deepnegative_v1_75t.pt"),
         "ng_deepnegative_v1_75t",
     ],
     [
-        "../../checkpoints/embedding/EasyNegativeV2.safetensors",
+        os.path.join(PROJECT_DIR, "checkpoints/embedding/EasyNegativeV2.safetensors"),
         "EasyNegativeV2",
     ],
     [
-        "../../checkpoints/embedding/bad_prompt_version2-neg.pt",
+        os.path.join(PROJECT_DIR, "checkpoints/embedding/bad_prompt_version2-neg.pt"),
         "bad_prompt_version2-neg",
     ],
 ]
@@ -466,7 +476,7 @@ if referencenet_model_name is not None and not use_v2v_predictor:
     referencenet = load_referencenet_by_name(
         model_name=referencenet_model_name,
         # sd_model=sd_model_path,
-        # sd_model="../../checkpoints//Moore-AnimateAnyone/AnimateAnyone/reference_unet.pth",
+        # sd_model=os.path.join(PROJECT_DIR, "checkpoints//Moore-AnimateAnyone/AnimateAnyone/reference_unet.pth"),
         sd_referencenet_model=referencenet_model_path,
         cross_attention_dim=cross_attention_dim,
     )
@@ -517,7 +527,7 @@ for model_name, sd_model_params in sd_model_params_dict.items():
             model_name=unet_model_name,
             sd_unet_model=unet_model_path,
             sd_model=sd_model_path,
-            # sd_model="../../checkpoints//Moore-AnimateAnyone/AnimateAnyone/denoising_unet.pth",
+            # sd_model=os.path.join(PROJECT_DIR, "checkpoints//Moore-AnimateAnyone/AnimateAnyone/denoising_unet.pth"),
             cross_attention_dim=cross_attention_dim,
             need_t2i_facein=facein_model_name is not None,
             # facein 目前没参与训练，但在unet中定义了，载入相关参数会报错，所以用strict控制

MuseV/scripts/gradio/gradio_video2video.py

@@ -66,7 +66,7 @@ args_dict = {
     "enable_zero_snr": False,
     "end_to_end": True,
     "face_image_path": None,
-    "facein_model_cfg_path": "../../configs/model/facein.py",
+    "facein_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/facein.py"),
     "facein_model_name": None,
     "facein_scale": 1.0,
     "fix_condition_images": False,
@@ -79,14 +79,18 @@ args_dict = {
     "img_length_ratio": 1.0,
     "img_weight": 0.001,
     "interpolation_factor": 1,
-    "ip_adapter_face_model_cfg_path": "../../configs/model/ip_adapter.py",
+    "ip_adapter_face_model_cfg_path": os.path.join(
+        PROJECT_DIR, "configs/model/ip_adapter.py"
+    ),
     "ip_adapter_face_model_name": None,
     "ip_adapter_face_scale": 1.0,
-    "ip_adapter_model_cfg_path": "../../configs/model/ip_adapter.py",
-    "ip_adapter_model_name": "musev_referencenet_pose",
+    "ip_adapter_model_cfg_path": os.path.join(
+        PROJECT_DIR, "configs/model/ip_adapter.py"
+    ),
+    "ip_adapter_model_name": "musev_referencenet",
     "ip_adapter_scale": 1.0,
     "ipadapter_image_path": None,
-    "lcm_model_cfg_path": "../../configs/model/lcm_model.py",
+    "lcm_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/lcm_model.py"),
     "lcm_model_name": None,
     "log_level": "INFO",
     "motion_speed": 8.0,
@@ -100,7 +104,7 @@ args_dict = {
     "need_return_videos": False,
     "need_video2video": False,
     "negative_prompt": "V2",
-    "negprompt_cfg_path": "../../configs/model/negative_prompt.py",
+    "negprompt_cfg_path": os.path.join(PROJECT_DIR, "configs/model/negative_prompt.py"),
     "noise_type": "video_fusion",
     "num_inference_steps": 30,
     "output_dir": "./results/",
@@ -115,22 +119,26 @@ args_dict = {
     "redraw_condition_image_with_ipdapter": True,
     "redraw_condition_image_with_referencenet": True,
     "referencenet_image_path": None,
-    "referencenet_model_cfg_path": "../../configs/model/referencenet.py",
+    "referencenet_model_cfg_path": os.path.join(
+        PROJECT_DIR, "configs/model/referencenet.py"
+    ),
     "referencenet_model_name": "musev_referencenet",
     "sample_rate": 1,
     "save_filetype": "mp4",
     "save_images": False,
-    "sd_model_cfg_path": "../../configs/model/T2I_all_model.py",
+    "sd_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/T2I_all_model.py"),
     "sd_model_name": "majicmixRealv6Fp16",
     "seed": None,
     "strength": 0.8,
     "target_datas": "boy_dance2",
-    "test_data_path": "./configs/infer/testcase_video_famous.yaml",
+    "test_data_path": os.path.join(
+        PROJECT_DIR, "configs/infer/testcase_video_famous.yaml"
+    ),
     "time_size": 12,
-    "unet_model_cfg_path": "../../configs/model/motion_model.py",
+    "unet_model_cfg_path": os.path.join(PROJECT_DIR, "configs/model/motion_model.py"),
     "unet_model_name": "musev_referencenet_pose",
     "use_condition_image": True,
-    "vae_model_path": "../../checkpoints/vae/sd-vae-ft-mse",
+    "vae_model_path": os.path.join(PROJECT_DIR, "checkpoints/vae/sd-vae-ft-mse"),
     "video_guidance_scale": 3.5,
     "video_guidance_scale_end": None,
     "video_guidance_scale_method": "linear",
@@ -141,7 +149,9 @@ args_dict = {
     "video_overlap": 1,
     "video_strength": 1.0,
     "vision_clip_extractor_class_name": "ImageClipVisionFeatureExtractor",
-    "vision_clip_model_path": "../../checkpoints/IP-Adapter/models/image_encoder",
+    "vision_clip_model_path": os.path.join(
+        PROJECT_DIR, "checkpoints/IP-Adapter/models/image_encoder"
+    ),
     "w_ind_noise": 0.5,
     "which2video": "video_middle",
     "width": None,
@@ -279,17 +289,17 @@ else:
     )
 b = 1
 negative_embedding = [
-    ["../../checkpoints/embedding/badhandv4.pt", "badhandv4"],
+    [os.path.join(PROJECT_DIR, "checkpoints/embedding/badhandv4.pt"), "badhandv4"],
     [
-        "../../checkpoints/embedding/ng_deepnegative_v1_75t.pt",
+        os.path.join(PROJECT_DIR, "checkpoints/embedding/ng_deepnegative_v1_75t.pt"),
         "ng_deepnegative_v1_75t",
     ],
     [
-        "../../checkpoints/embedding/EasyNegativeV2.safetensors",
+        os.path.join(PROJECT_DIR, "checkpoints/embedding/EasyNegativeV2.safetensors"),
         "EasyNegativeV2",
     ],
     [
-        "../../checkpoints/embedding/bad_prompt_version2-neg.pt",
+        os.path.join(PROJECT_DIR, "checkpoints/embedding/bad_prompt_version2-neg.pt"),
         "bad_prompt_version2-neg",
     ],
 ]
@@ -494,7 +504,7 @@ if referencenet_model_name is not None:
     referencenet = load_referencenet_by_name(
         model_name=referencenet_model_name,
         # sd_model=sd_model_path,
-        # sd_model="../../checkpoints/Moore-AnimateAnyone/AnimateAnyone/reference_unet.pth",
+        # sd_model=os.path.join(PROJECT_DIR, "checkpoints//Moore-AnimateAnyone/AnimateAnyone/reference_unet.pth"),
         sd_referencenet_model=referencenet_model_path,
         cross_attention_dim=cross_attention_dim,
     )
@@ -554,7 +564,7 @@ for model_name, sd_model_params in sd_model_params_dict.items():
         model_name=unet_model_name,
         sd_unet_model=unet_model_path,
         sd_model=sd_model_path,
-        # sd_model="../../checkpoints/Moore-AnimateAnyone/AnimateAnyone/denoising_unet.pth",
+        # sd_model=os.path.join(PROJECT_DIR, "checkpoints//Moore-AnimateAnyone/AnimateAnyone/denoising_unet.pth"),
         cross_attention_dim=cross_attention_dim,
         need_t2i_facein=facein_model_name is not None,
         # facein 目前没参与训练，但在unet中定义了，载入相关参数会报错，所以用strict控制

app_gradio_space.py

@@ -2,6 +2,7 @@ import os
 import time
 import pdb
 import torch
+
 torch.jit.script = lambda f: f
 import timm
 import cuid
@@ -20,6 +21,7 @@ CheckpointsDir = os.path.join(ProjectDir, "checkpoints")
 ignore_video2video = False
 max_image_edge = 960
 
+sys.path.insert(0, f"{ProjectDir}")
 sys.path.insert(0, f"{ProjectDir}/MMCM")
 sys.path.insert(0, f"{ProjectDir}/scripts/gradio")