Refactor Config System (#338)

configs/seko_talk/seko_talk_16_fp8_dist_compile.json

@@ -8,7 +8,6 @@
     "self_attn_1_type": "sage_attn2",
     "cross_attn_1_type": "sage_attn2",
     "cross_attn_2_type": "sage_attn2",
-    "seed": 42,
     "sample_guide_scale": 1.0,
     "sample_shift": 5,
     "enable_cfg": false,
@@ -26,5 +25,46 @@
     "t5_quantized": true,
     "t5_quant_scheme": "fp8",
     "compile": true,
-    "compile_shapes": [[480, 832], [544, 960], [720, 1280], [832, 480], [960, 544], [1280, 720], [480, 480], [576, 576], [704, 704], [960, 960]]
+    "compile_shapes": [
+        [
+            480,
+            832
+        ],
+        [
+            544,
+            960
+        ],
+        [
+            720,
+            1280
+        ],
+        [
+            832,
+            480
+        ],
+        [
+            960,
+            544
+        ],
+        [
+            1280,
+            720
+        ],
+        [
+            480,
+            480
+        ],
+        [
+            576,
+            576
+        ],
+        [
+            704,
+            704
+        ],
+        [
+            960,
+            960
+        ]
+    ]
 }

configs/skyreels/wan_skyreels_v2_df.json

@@ -12,7 +12,6 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
     "sample_guide_scale": 5,
     "sample_shift": 3
 }

configs/skyreels/wan_skyreels_v2_i2v.json

@@ -7,7 +7,6 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
     "sample_guide_scale": 5,
     "sample_shift": 3
 }

configs/skyreels/wan_skyreels_v2_t2v.json

@@ -7,7 +7,6 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
     "sample_guide_scale": 6,
     "sample_shift": 8
 }

configs/video_frame_interpolation/wan_t2v.json

@@ -7,7 +7,6 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
     "sample_guide_scale": 6,
     "sample_shift": 8,
     "enable_cfg": true,

configs/wan/wan_flf2v.json

@@ -6,7 +6,6 @@
     "self_attn_1_type": "sage_attn2",
     "cross_attn_1_type": "sage_attn2",
     "cross_attn_2_type": "sage_attn2",
-    "seed": 442,
     "sample_guide_scale": 5,
     "sample_shift": 16,
     "enable_cfg": true,

configs/wan/wan_i2v.json

@@ -6,7 +6,6 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 442,
     "sample_guide_scale": 5,
     "sample_shift": 3,
     "enable_cfg": true,

configs/wan/wan_t2v.json

@@ -7,7 +7,6 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
     "sample_guide_scale": 6,
     "sample_shift": 8,
     "enable_cfg": true,

configs/wan/wan_t2v_enhancer.json

@@ -7,12 +7,13 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
     "sample_guide_scale": 6,
     "sample_shift": 8,
     "enable_cfg": true,
     "cpu_offload": false,
     "sub_servers": {
-        "prompt_enhancer": ["http://localhost:9001"]
+        "prompt_enhancer": [
+            "http://localhost:9001"
+        ]
     }
 }

configs/wan/wan_vace.json

@@ -6,7 +6,6 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
     "sample_guide_scale": 5,
     "sample_shift": 16,
     "enable_cfg": true,

configs/wan22/wan_animate.json

@@ -8,7 +8,6 @@
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
     "adapter_attn_type": "flash_attn3",
-    "seed": 42,
     "sample_shift": 5.0,
     "sample_guide_scale": 5.0,
     "enable_cfg": false,

configs/wan22/wan_animate_replace.json

@@ -8,7 +8,6 @@
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
     "adapter_attn_type": "flash_attn3",
-    "seed": 42,
     "sample_shift": 5.0,
     "sample_guide_scale": 5.0,
     "enable_cfg": false,

configs/wan22/wan_distill_moe_flf2v.json

@@ -7,8 +7,10 @@
     "self_attn_1_type": "sage_attn2",
     "cross_attn_1_type": "sage_attn2",
     "cross_attn_2_type": "sage_attn2",
-    "seed": 42,
-    "sample_guide_scale": [3.5, 3.5],
+    "sample_guide_scale": [
+        3.5,
+        3.5
+    ],
     "sample_shift": 16,
     "enable_cfg": false,
     "cpu_offload": true,
@@ -17,5 +19,10 @@
     "vae_cpu_offload": false,
     "use_image_encoder": false,
     "boundary_step_index": 2,
-    "denoising_step_list": [1000, 750, 500, 250]
+    "denoising_step_list": [
+        1000,
+        750,
+        500,
+        250
+    ]
 }

configs/wan22/wan_distill_moe_flf2v_fp8.json

@@ -7,8 +7,10 @@
     "self_attn_1_type": "sage_attn2",
     "cross_attn_1_type": "sage_attn2",
     "cross_attn_2_type": "sage_attn2",
-    "seed": 42,
-    "sample_guide_scale": [3.5, 3.5],
+    "sample_guide_scale": [
+        3.5,
+        3.5
+    ],
     "sample_shift": 16,
     "enable_cfg": false,
     "cpu_offload": true,
@@ -17,7 +19,12 @@
     "vae_cpu_offload": false,
     "use_image_encoder": false,
     "boundary_step_index": 2,
-    "denoising_step_list": [1000, 750, 500, 250],
+    "denoising_step_list": [
+        1000,
+        750,
+        500,
+        250
+    ],
     "mm_config": {
         "mm_type": "W-fp8-channel-sym-A-fp8-channel-sym-dynamic-Sgl"
     },

configs/wan22/wan_distill_moe_flf2v_int8.json

@@ -7,8 +7,10 @@
     "self_attn_1_type": "sage_attn2",
     "cross_attn_1_type": "sage_attn2",
     "cross_attn_2_type": "sage_attn2",
-    "seed": 42,
-    "sample_guide_scale": [3.5, 3.5],
+    "sample_guide_scale": [
+        3.5,
+        3.5
+    ],
     "sample_shift": 16,
     "enable_cfg": false,
     "cpu_offload": true,
@@ -17,7 +19,12 @@
     "vae_cpu_offload": false,
     "use_image_encoder": false,
     "boundary_step_index": 2,
-    "denoising_step_list": [1000, 750, 500, 250],
+    "denoising_step_list": [
+        1000,
+        750,
+        500,
+        250
+    ],
     "mm_config": {
         "mm_type": "W-int8-channel-sym-A-int8-channel-sym-dynamic-Vllm"
     },

configs/wan22/wan_distill_moe_flf2v_with_lora.json

@@ -7,8 +7,10 @@
     "self_attn_1_type": "sage_attn2",
     "cross_attn_1_type": "sage_attn2",
     "cross_attn_2_type": "sage_attn2",
-    "seed": 42,
-    "sample_guide_scale": [3.5, 3.5],
+    "sample_guide_scale": [
+        3.5,
+        3.5
+    ],
     "sample_shift": 16,
     "enable_cfg": false,
     "cpu_offload": true,
@@ -17,7 +19,12 @@
     "vae_cpu_offload": false,
     "use_image_encoder": false,
     "boundary_step_index": 2,
-    "denoising_step_list": [1000, 750, 500, 250],
+    "denoising_step_list": [
+        1000,
+        750,
+        500,
+        250
+    ],
     "lora_configs": [
         {
             "name": "low_noise_model",

configs/wan22/wan_moe_flf2v.json

@@ -6,11 +6,13 @@
     "self_attn_1_type": "sage_attn2",
     "cross_attn_1_type": "sage_attn2",
     "cross_attn_2_type": "sage_attn2",
-    "seed": 42,
     "sample_shift": 16,
     "enable_cfg": true,
     "cpu_offload": true,
-    "sample_guide_scale": [3.5, 3.5],
+    "sample_guide_scale": [
+        3.5,
+        3.5
+    ],
     "offload_granularity": "model",
     "t5_cpu_offload": false,
     "vae_cpu_offload": false,

configs/wan22/wan_moe_i2v.json

@@ -7,8 +7,10 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
-    "sample_guide_scale": [3.5, 3.5],
+    "sample_guide_scale": [
+        3.5,
+        3.5
+    ],
     "sample_shift": 5.0,
     "enable_cfg": true,
     "cpu_offload": true,

configs/wan22/wan_moe_i2v_4090.json

@@ -7,8 +7,10 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
-    "sample_guide_scale": [3.5, 3.5],
+    "sample_guide_scale": [
+        3.5,
+        3.5
+    ],
     "sample_shift": 5.0,
     "enable_cfg": false,
     "cpu_offload": true,
@@ -16,5 +18,10 @@
     "boundary": 0.900,
     "use_image_encoder": false,
     "boundary_step_index": 2,
-    "denoising_step_list": [1000, 750, 500, 250]
+    "denoising_step_list": [
+        1000,
+        750,
+        500,
+        250
+    ]
 }

configs/wan22/wan_moe_i2v_audio.json

@@ -10,8 +10,10 @@
     "self_attn_1_type": "flash_attn3",
     "cross_attn_1_type": "flash_attn3",
     "cross_attn_2_type": "flash_attn3",
-    "seed": 42,
-    "sample_guide_scale": [1.0, 1.0],
+    "sample_guide_scale": [
+        1.0,
+        1.0
+    ],
     "sample_shift": 5.0,
     "enable_cfg": false,
     "cpu_offload": true,
@@ -23,14 +25,14 @@
     "use_31_block": false,
     "lora_configs": [
         {
-        "name": "high_noise_model",
-        "path": "/mnt/Text2Video/wuzhuguanyu/Wan21_T2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors",
-        "strength": 1.0
+            "name": "high_noise_model",
+            "path": "/mnt/Text2Video/wuzhuguanyu/Wan21_T2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors",
+            "strength": 1.0
         },
         {
-        "name": "low_noise_model",
-        "path": "/mnt/Text2Video/wuzhuguanyu/Wan21_T2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors",
-        "strength": 1.0
+            "name": "low_noise_model",
+            "path": "/mnt/Text2Video/wuzhuguanyu/Wan21_T2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors",
+            "strength": 1.0
         }
-      ]
+    ]
 }