diff --git "a/run.log" "b/run.log"
new file mode 100644--- /dev/null
+++ "b/run.log"
@@ -0,0 +1,706 @@
+01/27/2026 20:01:27 - INFO - __main__ - Conducting Command: t2v/scripts/calib_wan_i2v.py --model_path Wan-AI/Wan2.2-I2V-A14B-Diffusers --calib_data calib_data_wan_i2v/ --calib_config t2v/configs/quant/wan/W4A4_i2v_nvfp4.yaml --outdir quant_output/ --transformer 1
+01/27/2026 20:01:27 - INFO - __main__ - Loading calibration data from calib_data_wan_i2v/...
+01/27/2026 20:01:27 - INFO - __main__ - Transformer mode: 1
+01/27/2026 20:02:01 - INFO - __main__ - T2 calib data: 74 timesteps
+01/27/2026 20:02:01 - INFO - __main__ - Skipping transformer_1 (--transformer=2)
+01/27/2026 20:02:01 - I01/27/2026 20:02:05 - INFO - __main__ - Selected 10 steps from 26 available
+01/2701/27/2026 20:02:35 - INFO - __main__ - Selected 10 steps from 74 available
+01/27/2026 20:02:35 - INFO - __main__ - 
+============================================================
+01/27/2026 20:02:35 - INFO - __main__ - Calibrating Transformer 2
+01/27/2026 20:02:35 - INFO - __main__ - ============================================================
+01/27/2026 20:02:35 - INFO - qdiff.models.quant_model - 
+ --------------- refactoring quant layers --------------- 
+
+01/27/2026 20:02:38 - INFO - qdiff.models.quant_model - Set skip_lora_during_calib=True for all QuantLayers
+01/27/2026 20:02:38 - INFO - __main__ - T2 Calibration data shapes:
+01/27/2026 20:02:38 - INFO - __main__ -   xs: torch.Size([60, 36, 21, 90, 160])
+01/27/2026 20:02:38 - INFO - __main__ -   ts: torch.Size([60])
+01/27/2026 20:02:38 - INFO - __main__ -   conds: torch.Size([60, 512, 4096])
+01/27/2026 20:02:38 - INFO - __main__ - T2: Calculating activation statistics for smooth quantization...
+01/27/2026 20:02:38 - INFO - __main__ - T2: Smooth quant timeranges: [[0, 1000]]
+01/27/2026 20:02:38 - INFO - __main__ - T2: Calibration timestep coverage by timeran01/27/2026 20:21:22 - INFO - __main__ - T1: Smooth quant act_scale initialized in01/27/2026 20:21:51 - INFO - __main__ - T2: Smooth quant act_scale initialized in 86/86 layers
+01/27/2026 20:21:51 - INFO - __main__ - T2: Timerange 0: All layers have valid act_scale01/27/2026 20:22:01 - INFO - __main__ - T1: Weight initialization done!
+01/27/20201/27/2026 20:22:30 - INFO - __main__ - T2: Weight initialization done!
+01/27/2026 20:22:30 - INFO - __main__ - T2: Weight delta initialized in 406/86 layers
+01/27/2026 20:22:30 - INFO - __main__ - T2: Initializing activation quantization...
+01/27/2026 20:22:30 - INFO - __main__ - T2: Using dynamic quantization
+01/27/2026 20:22:30 - INFO - __main__ - T2: Activation initialization done!
+01/27/2026 20:22:30 - INFO - __main__ - T2: Initializing TQE frame weights (M)...
+01/27/2026 20:22:30 - INFO - __main__ - T2: Detected 21 frames for TQE M initialization
+01/27/2026 20:22:30 - INFO - __main__ - T2: Found 320 layers for TQE M initialization
+01/27/2026 20:22:30 - INFO - __main__ - T2: Layer filter: attention
+01/201/27/2026 20:25:06 - INFO - __main__ - T1: Processing layer batch 2/7 (layers 5001/2701/27/2026 20:28:16 - INFO - __main__ - T1: Processing layer batch 3/7 (layers 1001/27/201/27/2026 20:31:22 - INFO - __main__ - T1: Processing layer batch 4/7 (layers 1501/27/201/27/2026 20:34:34 - INFO - __main__ - T1: Processing layer batch 5/7 (layers 2001/27/201/27/2026 20:37:40 - INFO - __main__ - T1: Processing layer batch 6/7 (layers 2501/27/201/27/2026 20:40:46 - INFO - __main__ - T1: Processing layer batch 7/7 (layers 3001/27/201/27/2026 20:42:46 - INFO - __main__ - T1: TQE M initialization complete! Initia01/27/2026 20:43:02 - INFO - __main__ - T2: TQE M initialization complete! Initialized 240 layers
+01/27/2026 20:43:02 - INFO - qdiff.models.quant_model - Set skip_lora_during_calib=False for all QuantLayers
+01/27/2026 20:43:02 - INFO - __main__ - T2: Starting Q-VDiT optimization phase...
+01/27/2026 20:43:02 - INFO - __main__ -   Weight optimization: True
+01/27/2026 20:43:02 - INFO - __main__ -   Activation optimization: False
+01/27/2026 20:43:03 - INFO - __main__ - T2: Optimizing weight params: ['delta']
+01/27/2026 20:43:03 - INFO - __main__ - T2: Running 1000 optimization iterations...
+01/27/2026 20:43:03 - INFO - __main__ - T2: Loss type: relation
+01/27/2026 20:43:03 - INFO - qdiff.optimization.model_recon - WAN per-block reconstruction: 40 blocks
+01/27/2026 20:43:03 - INFO - qdiff.optimization.model_recon - This approach processes one block at a time to avoid OOM
+01/27/2026 20:43:03 - INFO - qdiff.optimization.model_recon - Processing WAN block 0/39
+01/27/2026 20:43:03 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/27/2026 20:43:03 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 0...
+01/27/2026 20:43:17 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/27/2026 20:43:17 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/27/2026 20:43:17 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.57GB
+01/27/2026 20:4301/27/2026 20:45:09 - INFO - qdiff.optimization.model_recon -   Block 0 iter 50/101/27/2026 20:45:01/27/2026 20:47:16 - INFO - qdiff.optimization.model_recon -   Block 0 iter 100/01/27/2026 20:47:301/27/2026 20:49:23 - INFO - qdiff.optimization.model_recon -   Block 0 iter 150/01/27/2026 20:49:301/27/2026 20:51:29 - INFO - qdiff.optimization.model_recon -   Block 0 iter 200/01/27/2026 20:51:401/27/2026 20:53:33 - INFO - qdiff.optimization.model_recon -   Block 0 iter 250/01/27/2026 20:53:501/27/2026 20:55:37 - INFO - qdiff.optimization.model_recon -   Block 0 iter 300/01/27/2026 20:55:501/27/2026 20:57:45 - INFO - qdiff.optimization.model_recon -   Block 0 iter 350/01/27/2026 20:58:001/27/2026 20:59:49 - INFO - qdiff.optimization.model_recon -   Block 0 iter 400/01/27/2026 21:00:001/27/2026 21:01:54 - INFO - qdiff.optimization.model_recon -   Block 0 iter 450/01/27/2026 21:02:101/27/2026 21:03:57 - INFO - qdiff.optimization.model_recon -   Block 0 iter 500/01/27/2026 21:04:201/27/2026 21:06:00 - INFO - qdiff.optimization.model_recon -   Block 0 iter 550/01/27/2026 21:06:201/27/2026 21:08:02 - INFO - qdiff.optimization.model_recon -   Block 0 iter 600/01/27/2026 21:08:301/27/2026 21:10:05 - INFO - qdiff.optimization.model_recon -   Block 0 iter 650/01/27/2026 21:10:401/27/2026 21:12:07 - INFO - qdiff.optimization.model_recon -   Block 0 iter 700/01/27/2026 21:12:401/27/2026 21:14:09 - INFO - qdiff.optimization.model_recon -   Block 0 iter 750/01/27/2026 21:14:401/27/2026 21:16:09 - INFO - qdiff.optimization.model_recon -   Block 0 iter 800/01/27/2026 21:16:501/27/2026 21:18:10 - INFO - qdiff.optimization.model_recon -   Block 0 iter 850/01/27/2026 21:18:501/27/2026 21:20:13 - INFO - qdiff.optimization.model_recon -   Block 0 iter 900/01/27/2026 21:20:501/27/2026 21:22:14 - INFO - qdiff.optimization.model_recon -   Block 0 iter 950/01/27/2026 21:22:501/27/2026 21:24:14 - INFO - qdiff.optimization.model_recon -   Block 0 optimizat01/27/2026 21:25:00 - INFO - qdiff.optimization.model_recon -   Block 0 optimization complete
+01/27/2026 21:25:00 - INFO - qdiff.optimization.model_recon - Processing WAN block 1/39
+01/27/2026 21:25:00 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/27/2026 21:25:00 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 1...
+01/27/2026 21:25:18 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/27/2026 21:25:18 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/27/2026 21:25:19 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.62GB
+01/27/2026 21:2501/27/2026 21:26:38 - INFO - qdiff.optimization.model_recon -   Block 1 iter 50/101/27/2026 21:27:01/27/2026 21:28:44 - INFO - qdiff.optimization.model_recon -   Block 1 iter 100/01/27/2026 21:29:301/27/2026 21:30:51 - INFO - qdiff.optimization.model_recon -   Block 1 iter 150/01/27/2026 21:31:301/27/2026 21:32:58 - INFO - qdiff.optimization.model_recon -   Block 1 iter 200/01/27/2026 21:33:401/27/2026 21:35:09 - INFO - qdiff.optimization.model_recon -   Block 1 iter 250/01/27/2026 21:35:501/27/2026 21:37:21 - INFO - qdiff.optimization.model_recon -   Block 1 iter 300/01/27/2026 21:38:001/27/2026 21:39:25 - INFO - qdiff.optimization.model_recon -   Block 1 iter 350/01/27/2026 21:40:101/27/2026 21:41:29 - INFO - qdiff.optimization.model_recon -   Block 1 iter 400/01/27/2026 21:42:101/27/2026 21:43:31 - INFO - qdiff.optimization.model_recon -   Block 1 iter 450/01/27/2026 21:44:201/27/2026 21:45:34 - INFO - qdiff.optimization.model_recon -   Block 1 iter 500/01/27/2026 21:46:201/27/2026 21:47:40 - INFO - qdiff.optimization.model_recon -   Block 1 iter 550/01/27/2026 21:48:201/27/2026 21:49:45 - INFO - qdiff.optimization.model_recon -   Block 1 iter 600/01/27/2026 21:50:301/27/2026 21:51:52 - INFO - qdiff.optimization.model_recon -   Block 1 iter 650/01/27/2026 21:52:301/27/2026 21:53:59 - INFO - qdiff.optimization.model_recon -   Block 1 iter 700/01/27/2026 21:54:401/27/2026 21:56:06 - INFO - qdiff.optimization.model_recon -   Block 1 iter 750/01/27/2026 21:56:401/27/2026 21:58:11 - INFO - qdiff.optimization.model_recon -   Block 1 iter 800/01/27/2026 21:58:501/27/2026 22:00:10 - INFO - qdiff.optimization.model_recon -   Block 1 iter 850/01/27/2026 22:00:501/27/2026 22:02:08 - INFO - qdiff.optimization.model_recon -   Block 1 iter 900/01/27/2026 22:02:501/27/2026 22:04:12 - INFO - qdiff.optimization.model_recon -   Block 1 iter 950/01/27/2026 22:05:001/27/2026 22:06:15 - INFO - qdiff.optimization.model_recon -   Block 1 optimizat01/27/2026 22:07:03 - INFO - qdiff.optimization.model_recon -   Block 1 optimization complete
+01/27/2026 22:07:03 - INFO - qdiff.optimization.model_recon - Processing WAN block 2/39
+01/27/2026 22:07:03 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/27/2026 22:07:03 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 2...
+01/27/2026 22:07:24 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/27/2026 22:07:24 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/27/2026 22:07:24 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.64GB
+01/27/2026 22:0701/27/2026 22:08:42 - INFO - qdiff.optimization.model_recon -   Block 2 iter 50/101/27/2026 22:09:01/27/2026 22:10:49 - INFO - qdiff.optimization.model_recon -   Block 2 iter 100/01/27/2026 22:11:301/27/2026 22:12:54 - INFO - qdiff.optimization.model_recon -   Block 2 iter 150/01/27/2026 22:13:301/27/2026 22:14:56 - INFO - qdiff.optimization.model_recon -   Block 2 iter 200/01/27/2026 22:15:401/27/2026 22:16:59 - INFO - qdiff.optimization.model_recon -   Block 2 iter 250/01/27/2026 22:17:401/27/2026 22:18:58 - INFO - qdiff.optimization.model_recon -   Block 2 iter 300/01/27/2026 22:19:401/27/2026 22:20:58 - INFO - qdiff.optimization.model_recon -   Block 2 iter 350/01/27/2026 22:21:401/27/2026 22:22:59 - INFO - qdiff.optimization.model_recon -   Block 2 iter 400/01/27/2026 22:23:401/27/2026 22:25:01 - INFO - qdiff.optimization.model_recon -   Block 2 iter 450/01/27/2026 22:25:401/27/2026 22:27:04 - INFO - qdiff.optimization.model_recon -   Block 2 iter 500/01/27/2026 22:27:401/27/2026 22:29:07 - INFO - qdiff.optimization.model_recon -   Block 2 iter 550/01/27/2026 22:29:401/27/2026 22:31:08 - INFO - qdiff.optimization.model_recon -   Block 2 iter 600/01/27/2026 22:31:401/27/2026 22:33:06 - INFO - qdiff.optimization.model_recon -   Block 2 iter 650/01/27/2026 22:33:401/27/2026 22:35:08 - INFO - qdiff.optimization.model_recon -   Block 2 iter 700/01/27/2026 22:35:401/27/2026 22:37:11 - INFO - qdiff.optimization.model_recon -   Block 2 iter 750/01/27/2026 22:37:401/27/2026 22:39:12 - INFO - qdiff.optimization.model_recon -   Block 2 iter 800/01/27/2026 22:39:401/27/2026 22:41:14 - INFO - qdiff.optimization.model_recon -   Block 2 iter 850/01/27/2026 22:41:401/27/2026 22:43:13 - INFO - qdiff.optimization.model_recon -   Block 2 iter 900/01/27/2026 22:43:401/27/2026 22:45:16 - INFO - qdiff.optimization.model_recon -   Block 2 iter 950/01/27/2026 22:45:401/27/2026 22:47:12 - INFO - qdiff.optimization.model_recon -   Block 2 optimizat01/27/2026 22:47:45 - INFO - qdiff.optimization.model_recon -   Block 2 optimization complete
+01/27/2026 22:47:45 - INFO - qdiff.optimization.model_recon - Processing WAN block 3/39
+01/27/2026 22:47:45 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/27/2026 22:47:45 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 3...
+01/27/2026 22:48:11 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/27/2026 22:48:11 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/27/2026 22:48:11 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.65GB
+01/27/2026 22:4801/27/2026 22:49:40 - INFO - qdiff.optimization.model_recon -   Block 3 iter 50/101/27/2026 22:50:01/27/2026 22:51:40 - INFO - qdiff.optimization.model_recon -   Block 3 iter 100/01/27/2026 22:52:101/27/2026 22:53:40 - INFO - qdiff.optimization.model_recon -   Block 3 iter 150/01/27/2026 22:54:101/27/2026 22:55:42 - INFO - qdiff.optimization.model_recon -   Block 3 iter 200/01/27/2026 22:56:101/27/2026 22:57:44 - INFO - qdiff.optimization.model_recon -   Block 3 iter 250/01/27/2026 22:58:101/27/2026 22:59:43 - INFO - qdiff.optimization.model_recon -   Block 3 iter 300/01/27/2026 23:00:101/27/2026 23:01:42 - INFO - qdiff.optimization.model_recon -   Block 3 iter 350/01/27/2026 23:02:101/27/2026 23:03:41 - INFO - qdiff.optimization.model_recon -   Block 3 iter 400/01/27/2026 23:04:001/27/2026 23:05:39 - INFO - qdiff.optimization.model_recon -   Block 3 iter 450/01/27/2026 23:06:001/27/2026 23:07:39 - INFO - qdiff.optimization.model_recon -   Block 3 iter 500/01/27/2026 23:08:001/27/2026 23:09:38 - INFO - qdiff.optimization.model_recon -   Block 3 iter 550/01/27/2026 23:10:001/27/2026 23:11:37 - INFO - qdiff.optimization.model_recon -   Block 3 iter 600/01/27/2026 23:12:001/27/2026 23:13:38 - INFO - qdiff.optimization.model_recon -   Block 3 iter 650/01/27/2026 23:14:001/27/2026 23:15:39 - INFO - qdiff.optimization.model_recon -   Block 3 iter 700/01/27/2026 23:16:001/27/2026 23:17:37 - INFO - qdiff.optimization.model_recon -   Block 3 iter 750/01/27/2026 23:18:001/27/2026 23:19:37 - INFO - qdiff.optimization.model_recon -   Block 3 iter 800/01/27/2026 23:19:501/27/2026 23:21:37 - INFO - qdiff.optimization.model_recon -   Block 3 iter 850/01/27/2026 23:21:501/27/2026 23:23:36 - INFO - qdiff.optimization.model_recon -   Block 3 iter 900/01/27/2026 23:23:501/27/2026 23:25:36 - INFO - qdiff.optimization.model_recon -   Block 3 iter 950/01/27/2026 23:25:501/27/2026 23:27:33 - INFO - qdiff.optimization.model_recon -   Block 3 optimizat01/27/2026 23:27:52 - INFO - qdiff.optimization.model_recon -   Block 3 optimization complete
+01/27/2026 23:27:52 - INFO - qdiff.optimization.model_recon - Processing WAN block 4/39
+01/27/2026 23:27:52 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/27/2026 23:27:52 - IN01/27/2026 23:28:02 - INFO - qdiff.optimization.model_recon -   Block input shape01/27/2026 23:28:22 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/27/2026 23:28:22 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/27/2026 23:28:22 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.67GB
+01/27/2026 23:2801/27/2026 23:30:05 - INFO - qdiff.optimization.model_recon -   Block 4 iter 50/101/27/2026 23:30:01/27/2026 23:32:03 - INFO - qdiff.optimization.model_recon -   Block 4 iter 100/01/27/2026 23:32:101/27/2026 23:34:02 - INFO - qdiff.optimization.model_recon -   Block 4 iter 150/01/27/2026 23:34:101/27/2026 23:36:02 - INFO - qdiff.optimization.model_recon -   Block 4 iter 200/01/27/2026 23:36:101/27/2026 23:37:59 - INFO - qdiff.optimization.model_recon -   Block 4 iter 250/01/27/2026 23:38:101/27/2026 23:39:58 - INFO - qdiff.optimization.model_recon -   Block 4 iter 300/01/27/2026 23:40:001/27/2026 23:41:57 - INFO - qdiff.optimization.model_recon -   Block 4 iter 350/01/27/2026 23:42:001/27/2026 23:43:54 - INFO - qdiff.optimization.model_recon -   Block 4 iter 400/01/27/2026 23:44:001/27/2026 23:45:52 - INFO - qdiff.optimization.model_recon -   Block 4 iter 450/01/27/2026 23:45:501/27/2026 23:47:48 - INFO - qdiff.optimization.model_recon -   Block 4 iter 500/01/27/2026 23:47:501/27/2026 23:49:48 - INFO - qdiff.optimization.model_recon -   Block 4 iter 550/01/27/2026 23:49:501/27/2026 23:51:49 - INFO - qdiff.optimization.model_recon -   Block 4 iter 600/01/27/2026 23:51:501/27/2026 23:53:50 - INFO - qdiff.optimization.model_recon -   Block 4 iter 650/01/27/2026 23:53:501/27/2026 23:55:50 - INFO - qdiff.optimization.model_recon -   Block 4 iter 700/01/27/2026 23:55:501/27/2026 23:57:50 - INFO - qdiff.optimization.model_recon -   Block 4 iter 750/1000: GPU 27.71GB
+01/27/2026 23:59:48 - INFO - qdiff.optimization.model_recon -   Block 4 iter 800/1000: GPU 27.71GB
+01/28/2026 00:01:47 - INFO - qdiff.optimization.model_recon -   Block 4 iter 850/1000: GPU 27.71GB
+01/28/2026 00:03:46 - INFO - qdiff.optimization.model_recon -   Block 4 iter 900/1000: GPU 27.71GB
+01/28/2026 00:05:49 - INFO - qdiff.optimization.model_recon -   Block 4 iter 950/1000: GPU 27.71GB
+01/28/2026 00:07:48 - INFO - qdiff.optimization.model_recon -   Block 4 optimization complete
+01/28/2026 00:07:48 - INFO - qdiff.optimization.model_recon - Processing WAN block 5/39
+01/28/2026 00:07:48 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 00:07:48 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 5...
+01/28/2026 00:08:23 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 00:08:23 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 00:08:23 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.68GB
+01/28/2026 00:08:25 - INFO - qdiff.optimization.model_recon -   Block 5 iter 0/1000: GPU 27.73GB
+01/28/2026 00:10:20 - INFO - qdiff.optimization.model_recon -   Block 5 iter 50/1000: GPU 27.73GB
+01/28/2026 00:12:18 - INFO - qdiff.optimization.model_recon -   Block 5 iter 100/1000: GPU 27.73GB
+01/28/2026 00:14:17 - INFO - qdiff.optimization.model_recon -   Block 5 iter 150/1000: GPU 27.73GB
+01/28/2026 00:16:14 - INFO - qdiff.optimization.model_recon -   Block 5 iter 200/1000: GPU 27.73GB
+01/28/2026 00:18:10 - INFO - qdiff.optimization.model_recon -   Block 5 iter 250/1000: GPU 27.73GB
+01/28/2026 00:20:07 - INFO - qdiff.optimization.model_recon -   Block 5 iter 300/1000: GPU 27.73GB
+01/28/2026 00:22:03 - INFO - qdiff.optimization.model_recon -   Block 5 iter 350/1000: GPU 27.73GB
+01/28/2026 00:24:00 - INFO - qdiff.optimization.model_recon -   Block 5 iter 400/1000: GPU 27.73GB
+01/28/2026 00:25:55 - INFO - qdiff.optimization.model_recon -   Block 5 iter 450/1000: GPU 27.73GB
+01/28/2026 00:27:51 - INFO - qdiff.optimization.model_recon -   Block 5 iter 500/1000: GPU 27.73GB
+01/28/2026 00:29:46 - INFO - qdiff.optimization.model_recon -   Block 5 iter 550/1000: GPU 27.73GB
+01/28/2026 00:31:44 - INFO - qdiff.optimization.model_recon -   Block 5 iter 600/1000: GPU 27.73GB
+01/28/2026 00:33:42 - INFO - qdiff.optimization.model_recon -   Block 5 iter 650/1000: GPU 27.73GB
+01/28/2026 00:35:41 - INFO - qdiff.optimization.model_recon -   Block 5 iter 700/1000: GPU 27.73GB
+01/28/2026 00:37:38 - INFO - qdiff.optimization.model_recon -   Block 5 iter 750/1000: GPU 27.73GB
+01/28/2026 00:39:35 - INFO - qdiff.optimization.model_recon -   Block 5 iter 800/1000: GPU 27.73GB
+01/28/2026 00:41:35 - INFO - qdiff.optimization.model_recon -   Block 5 iter 850/1000: GPU 27.73GB
+01/28/2026 00:43:31 - INFO - qdiff.optimization.model_recon -   Block 5 iter 900/1000: GPU 27.73GB
+01/28/2026 00:45:28 - INFO - qdiff.optimization.model_recon -   Block 5 iter 950/1000: GPU 27.73GB
+01/28/2026 00:47:25 - INFO - qdiff.optimization.model_recon -   Block 5 optimization complete
+01/28/2026 00:47:25 - INFO - qdiff.optimization.model_recon - Processing WAN block 6/39
+01/28/2026 00:47:25 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 00:47:25 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 6...
+01/28/2026 00:48:02 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 00:48:02 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 00:48:02 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.70GB
+01/28/2026 00:48:04 - INFO - qdiff.optimization.model_recon -   Block 6 iter 0/1000: GPU 27.74GB
+01/28/2026 00:50:01 - INFO - qdiff.optimization.model_recon -   Block 6 iter 50/1000: GPU 27.74GB
+01/28/2026 00:51:59 - INFO - qdiff.optimization.model_recon -   Block 6 iter 100/1000: GPU 27.74GB
+01/28/2026 00:53:56 - INFO - qdiff.optimization.model_recon -   Block 6 iter 150/1000: GPU 27.74GB
+01/28/2026 00:55:52 - INFO - qdiff.optimization.model_recon -   Block 6 iter 200/1000: GPU 27.74GB
+01/28/2026 00:57:47 - INFO - qdiff.optimization.model_recon -   Block 6 iter 250/1000: GPU 27.74GB
+01/28/2026 00:59:43 - INFO - qdiff.optimization.model_recon -   Block 6 iter 300/1000: GPU 27.74GB
+01/28/2026 01:01:38 - INFO - qdiff.optimization.model_recon -   Block 6 iter 350/1000: GPU 27.74GB
+01/28/2026 01:03:35 - INFO - qdiff.optimization.model_recon -   Block 6 iter 400/1000: GPU 27.74GB
+01/28/2026 01:05:32 - INFO - qdiff.optimization.model_recon -   Block 6 iter 450/1000: GPU 27.74GB
+01/28/2026 01:07:32 - INFO - qdiff.optimization.model_recon -   Block 6 iter 500/1000: GPU 27.74GB
+01/28/2026 01:09:29 - INFO - qdiff.optimization.model_recon -   Block 6 iter 550/1000: GPU 27.74GB
+01/28/2026 01:11:26 - INFO - qdiff.optimization.model_recon -   Block 6 iter 600/1000: GPU 27.74GB
+01/28/2026 01:13:20 - INFO - qdiff.optimization.model_recon -   Block 6 iter 650/1000: GPU 27.74GB
+01/28/2026 01:15:13 - INFO - qdiff.optimization.model_recon -   Block 6 iter 700/1000: GPU 27.74GB
+01/28/2026 01:17:13 - INFO - qdiff.optimization.model_recon -   Block 6 iter 750/1000: GPU 27.74GB
+01/28/2026 01:19:07 - INFO - qdiff.optimization.model_recon -   Block 6 iter 800/1000: GPU 27.74GB
+01/28/2026 01:21:05 - INFO - qdiff.optimization.model_recon -   Block 6 iter 850/1000: GPU 27.74GB
+01/28/2026 01:23:02 - INFO - qdiff.optimization.model_recon -   Block 6 iter 900/1000: GPU 27.74GB
+01/28/2026 01:25:04 - INFO - qdiff.optimization.model_recon -   Block 6 iter 950/1000: GPU 27.74GB
+01/28/2026 01:27:00 - INFO - qdiff.optimization.model_recon -   Block 6 optimization complete
+01/28/2026 01:27:00 - INFO - qdiff.optimization.model_recon - Processing WAN block 7/39
+01/28/2026 01:27:00 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 01:27:00 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 7...
+01/28/2026 01:27:42 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 01:27:42 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 01:27:43 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.71GB
+01/28/2026 01:27:46 - INFO - qdiff.optimization.model_recon -   Block 7 iter 0/1000: GPU 27.76GB
+01/28/2026 01:29:41 - INFO - qdiff.optimization.model_recon -   Block 7 iter 50/1000: GPU 27.76GB
+01/28/2026 01:31:39 - INFO - qdiff.optimization.model_recon -   Block 7 iter 100/1000: GPU 27.76GB
+01/28/2026 01:33:33 - INFO - qdiff.optimization.model_recon -   Block 7 iter 150/1000: GPU 27.76GB
+01/28/2026 01:35:30 - INFO - qdiff.optimization.model_recon -   Block 7 iter 200/1000: GPU 27.76GB
+01/28/2026 01:37:25 - INFO - qdiff.optimization.model_recon -   Block 7 iter 250/1000: GPU 27.76GB
+01/28/2026 01:39:21 - INFO - qdiff.optimization.model_recon -   Block 7 iter 300/1000: GPU 27.76GB
+01/28/2026 01:41:16 - INFO - qdiff.optimization.model_recon -   Block 7 iter 350/1000: GPU 27.76GB
+01/28/2026 01:43:15 - INFO - qdiff.optimization.model_recon -   Block 7 iter 400/1000: GPU 27.76GB
+01/28/2026 01:45:13 - INFO - qdiff.optimization.model_recon -   Block 7 iter 450/1000: GPU 27.76GB
+01/28/2026 01:47:08 - INFO - qdiff.optimization.model_recon -   Block 7 iter 500/1000: GPU 27.76GB
+01/28/2026 01:49:04 - INFO - qdiff.optimization.model_recon -   Block 7 iter 550/1000: GPU 27.76GB
+01/28/2026 01:51:00 - INFO - qdiff.optimization.model_recon -   Block 7 iter 600/1000: GPU 27.76GB
+01/28/2026 01:52:58 - INFO - qdiff.optimization.model_recon -   Block 7 iter 650/1000: GPU 27.76GB
+01/28/2026 01:54:54 - INFO - qdiff.optimization.model_recon -   Block 7 iter 700/1000: GPU 27.76GB
+01/28/2026 01:56:48 - INFO - qdiff.optimization.model_recon -   Block 7 iter 750/1000: GPU 27.76GB
+01/28/2026 01:58:44 - INFO - qdiff.optimization.model_recon -   Block 7 iter 800/1000: GPU 27.76GB
+01/28/2026 02:00:43 - INFO - qdiff.optimization.model_recon -   Block 7 iter 850/1000: GPU 27.76GB
+01/28/2026 02:02:42 - INFO - qdiff.optimization.model_recon -   Block 7 iter 900/1000: GPU 27.76GB
+01/28/2026 02:04:40 - INFO - qdiff.optimization.model_recon -   Block 7 iter 950/1000: GPU 27.76GB
+01/28/2026 02:06:33 - INFO - qdiff.optimization.model_recon -   Block 7 optimization complete
+01/28/2026 02:06:33 - INFO - qdiff.optimization.model_recon - Processing WAN block 8/39
+01/28/2026 02:06:33 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 02:06:33 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 8...
+01/28/2026 02:07:19 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 02:07:19 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 02:07:19 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.73GB
+01/28/2026 02:07:21 - INFO - qdiff.optimization.model_recon -   Block 8 iter 0/1000: GPU 27.77GB
+01/28/2026 02:09:16 - INFO - qdiff.optimization.model_recon -   Block 8 iter 50/1000: GPU 27.77GB
+01/28/2026 02:11:11 - INFO - qdiff.optimization.model_recon -   Block 8 iter 100/1000: GPU 27.77GB
+01/28/2026 02:13:05 - INFO - qdiff.optimization.model_recon -   Block 8 iter 150/1000: GPU 27.77GB
+01/28/2026 02:14:58 - INFO - qdiff.optimization.model_recon -   Block 8 iter 200/1000: GPU 27.77GB
+01/28/2026 02:16:52 - INFO - qdiff.optimization.model_recon -   Block 8 iter 250/1000: GPU 27.77GB
+01/28/2026 02:18:45 - INFO - qdiff.optimization.model_recon -   Block 8 iter 300/1000: GPU 27.77GB
+01/28/2026 02:20:39 - INFO - qdiff.optimization.model_recon -   Block 8 iter 350/1000: GPU 27.77GB
+01/28/2026 02:22:32 - INFO - qdiff.optimization.model_recon -   Block 8 iter 400/1000: GPU 27.77GB
+01/28/2026 02:24:27 - INFO - qdiff.optimization.model_recon -   Block 8 iter 450/1000: GPU 27.77GB
+01/28/2026 02:26:21 - INFO - qdiff.optimization.model_recon -   Block 8 iter 500/1000: GPU 27.77GB
+01/28/2026 02:28:17 - INFO - qdiff.optimization.model_recon -   Block 8 iter 550/1000: GPU 27.77GB
+01/28/2026 02:30:12 - INFO - qdiff.optimization.model_recon -   Block 8 iter 600/1000: GPU 27.77GB
+01/28/2026 02:32:06 - INFO - qdiff.optimization.model_recon -   Block 8 iter 650/1000: GPU 27.77GB
+01/28/2026 02:34:03 - INFO - qdiff.optimization.model_recon -   Block 8 iter 700/1000: GPU 27.77GB
+01/28/2026 02:35:58 - INFO - qdiff.optimization.model_recon -   Block 8 iter 750/1000: GPU 27.77GB
+01/28/2026 02:37:52 - INFO - qdiff.optimization.model_recon -   Block 8 iter 800/1000: GPU 27.77GB
+01/28/2026 02:39:46 - INFO - qdiff.optimization.model_recon -   Block 8 iter 850/1000: GPU 27.77GB
+01/28/2026 02:41:42 - INFO - qdiff.optimization.model_recon -   Block 8 iter 900/1000: GPU 27.77GB
+01/28/2026 02:43:36 - INFO - qdiff.optimization.model_recon -   Block 8 iter 950/1000: GPU 27.77GB
+01/28/2026 02:45:29 - INFO - qdiff.optimization.model_recon -   Block 8 optimization complete
+01/28/2026 02:45:29 - INFO - qdiff.optimization.model_recon - Processing WAN block 9/39
+01/28/2026 02:45:29 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 02:45:29 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 9...
+01/28/2026 02:46:18 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 02:46:18 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 02:46:19 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.74GB
+01/28/2026 02:46:21 - INFO - qdiff.optimization.model_recon -   Block 9 iter 0/1000: GPU 27.79GB
+01/28/2026 02:48:14 - INFO - qdiff.optimization.model_recon -   Block 9 iter 50/1000: GPU 27.79GB
+01/28/2026 02:50:06 - INFO - qdiff.optimization.model_recon -   Block 9 iter 100/1000: GPU 27.79GB
+01/28/2026 02:51:58 - INFO - qdiff.optimization.model_recon -   Block 9 iter 150/1000: GPU 27.79GB
+01/28/2026 02:53:51 - INFO - qdiff.optimization.model_recon -   Block 9 iter 200/1000: GPU 27.79GB
+01/28/2026 02:55:42 - INFO - qdiff.optimization.model_recon -   Block 9 iter 250/1000: GPU 27.79GB
+01/28/2026 02:57:33 - INFO - qdiff.optimization.model_recon -   Block 9 iter 300/1000: GPU 27.79GB
+01/28/2026 02:59:26 - INFO - qdiff.optimization.model_recon -   Block 9 iter 350/1000: GPU 27.79GB
+01/28/2026 03:01:20 - INFO - qdiff.optimization.model_recon -   Block 9 iter 400/1000: GPU 27.79GB
+01/28/2026 03:03:13 - INFO - qdiff.optimization.model_recon -   Block 9 iter 450/1000: GPU 27.79GB
+01/28/2026 03:05:07 - INFO - qdiff.optimization.model_recon -   Block 9 iter 500/1000: GPU 27.79GB
+01/28/2026 03:07:01 - INFO - qdiff.optimization.model_recon -   Block 9 iter 550/1000: GPU 27.79GB
+01/28/2026 03:08:55 - INFO - qdiff.optimization.model_recon -   Block 9 iter 600/1000: GPU 27.79GB
+01/28/2026 03:10:48 - INFO - qdiff.optimization.model_recon -   Block 9 iter 650/1000: GPU 27.79GB
+01/28/2026 03:12:40 - INFO - qdiff.optimization.model_recon -   Block 9 iter 700/1000: GPU 27.79GB
+01/28/2026 03:14:31 - INFO - qdiff.optimization.model_recon -   Block 9 iter 750/1000: GPU 27.79GB
+01/28/2026 03:16:22 - INFO - qdiff.optimization.model_recon -   Block 9 iter 800/1000: GPU 27.79GB
+01/28/2026 03:18:12 - INFO - qdiff.optimization.model_recon -   Block 9 iter 850/1000: GPU 27.79GB
+01/28/2026 03:20:05 - INFO - qdiff.optimization.model_recon -   Block 9 iter 900/1000: GPU 27.79GB
+01/28/2026 03:21:59 - INFO - qdiff.optimization.model_recon -   Block 9 iter 950/1000: GPU 27.79GB
+01/28/2026 03:23:51 - INFO - qdiff.optimization.model_recon -   Block 9 optimization complete
+01/28/2026 03:23:51 - INFO - qdiff.optimization.model_recon - Completed 10/40 blocks
+01/28/2026 03:23:51 - INFO - qdiff.optimization.model_recon - Processing WAN block 10/39
+01/28/2026 03:23:51 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 03:23:51 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 10...
+01/28/2026 03:24:44 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 03:24:44 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 03:24:45 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.76GB
+01/28/2026 03:24:47 - INFO - qdiff.optimization.model_recon -   Block 10 iter 0/1000: GPU 27.80GB
+01/28/2026 03:26:42 - INFO - qdiff.optimization.model_recon -   Block 10 iter 50/1000: GPU 27.80GB
+01/28/2026 03:28:33 - INFO - qdiff.optimization.model_recon -   Block 10 iter 100/1000: GPU 27.80GB
+01/28/2026 03:30:25 - INFO - qdiff.optimization.model_recon -   Block 10 iter 150/1000: GPU 27.80GB
+01/28/2026 03:32:18 - INFO - qdiff.optimization.model_recon -   Block 10 iter 200/1000: GPU 27.80GB
+01/28/2026 03:34:12 - INFO - qdiff.optimization.model_recon -   Block 10 iter 250/1000: GPU 27.80GB
+01/28/2026 03:36:04 - INFO - qdiff.optimization.model_recon -   Block 10 iter 300/1000: GPU 27.80GB
+01/28/2026 03:37:55 - INFO - qdiff.optimization.model_recon -   Block 10 iter 350/1000: GPU 27.80GB
+01/28/2026 03:39:47 - INFO - qdiff.optimization.model_recon -   Block 10 iter 400/1000: GPU 27.80GB
+01/28/2026 03:41:40 - INFO - qdiff.optimization.model_recon -   Block 10 iter 450/1000: GPU 27.80GB
+01/28/2026 03:43:33 - INFO - qdiff.optimization.model_recon -   Block 10 iter 500/1000: GPU 27.80GB
+01/28/2026 03:45:27 - INFO - qdiff.optimization.model_recon -   Block 10 iter 550/1000: GPU 27.80GB
+01/28/2026 03:47:21 - INFO - qdiff.optimization.model_recon -   Block 10 iter 60001/28/2026 03:47:2201/28/2026 03:49:16 - INFO - qdiff.optimization.model_recon -   Block 10 iter 65001/28/2026 03:49:1801/28/2026 03:51:09 - INFO - qdiff.optimization.model_recon -   Block 10 iter 70001/28/2026 03:51:1401/28/2026 03:53:01 - INFO - qdiff.optimization.model_recon -   Block 10 iter 75001/28/2026 03:53:0601/28/2026 03:54:51 - INFO - qdiff.optimization.model_recon -   Block 10 iter 80001/28/2026 03:54:5801/28/2026 03:56:41 - INFO - qdiff.optimization.model_recon -   Block 10 iter 85001/28/2026 03:56:5001/28/2026 03:58:31 - INFO - qdiff.optimization.model_recon -   Block 10 iter 90001/28/2026 03:58:4201/28/2026 04:00:23 - INFO - qdiff.optimization.model_recon -   Block 10 iter 95001/28/2026 04:00:3601/28/2026 04:02:14 - INFO - qdiff.optimization.model_recon -   Block 10 optimiza01/28/2026 04:02:29 - INFO - qdiff.optimization.model_recon -   Block 10 optimization complete
+01/28/2026 04:02:29 - INFO - qdiff.optimization.model_recon - Processing WAN block 11/39
+01/28/2026 04:02:29 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 04:02:29 - INF01/28/2026 04:03:12 - INFO - qdiff.optimization.model_recon -   Block input shape01/28/2026 04:03:27 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 04:03:27 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 04:03:28 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.77GB
+01/28/2026 04:03:01/28/2026 04:05:09 - INFO - qdiff.optimization.model_recon -   Block 11 iter 50/01/28/2026 04:05:201/28/2026 04:07:01 - INFO - qdiff.optimization.model_recon -   Block 11 iter 10001/28/2026 04:07:2001/28/2026 04:08:53 - INFO - qdiff.optimization.model_recon -   Block 11 iter 15001/28/2026 04:09:1601/28/2026 04:10:46 - INFO - qdiff.optimization.model_recon -   Block 11 iter 20001/28/2026 04:11:1001/28/2026 04:12:37 - INFO - qdiff.optimization.model_recon -   Block 11 iter 25001/28/2026 04:13:0401/28/2026 04:14:30 - INFO - qdiff.optimization.model_recon -   Block 11 iter 30001/28/2026 04:14:5701/28/2026 04:16:20 - INFO - qdiff.optimization.model_recon -   Block 11 iter 35001/28/2026 04:16:5001/28/2026 04:18:12 - INFO - qdiff.optimization.model_recon -   Block 11 iter 40001/28/2026 04:18:4701/28/2026 04:20:05 - INFO - qdiff.optimization.model_recon -   Block 11 iter 45001/28/2026 04:20:4201/28/2026 04:21:57 - INFO - qdiff.optimization.model_recon -   Block 11 iter 50001/28/2026 04:22:3701/28/2026 04:23:50 - INFO - qdiff.optimization.model_recon -   Block 11 iter 55001/28/2026 04:24:3401/28/2026 04:25:46 - INFO - qdiff.optimization.model_recon -   Block 11 iter 60001/28/2026 04:26:2901/28/2026 04:27:39 - INFO - qdiff.optimization.model_recon -   Block 11 iter 65001/28/2026 04:28:2201/28/2026 04:29:32 - INFO - qdiff.optimization.model_recon -   Block 11 iter 70001/28/2026 04:30:1701/28/2026 04:31:25 - INFO - qdiff.optimization.model_recon -   Block 11 iter 75001/28/2026 04:32:1001/28/2026 04:33:16 - INFO - qdiff.optimization.model_recon -   Block 11 iter 80001/28/2026 04:34:0201/28/2026 04:35:08 - INFO - qdiff.optimization.model_recon -   Block 11 iter 85001/28/2026 04:35:5401/28/2026 04:36:58 - INFO - qdiff.optimization.model_recon -   Block 11 iter 90001/28/2026 04:37:4801/28/2026 04:38:51 - INFO - qdiff.optimization.model_recon -   Block 11 iter 95001/28/2026 04:39:4301/28/2026 04:40:42 - INFO - qdiff.optimization.model_recon -   Block 11 optimiza01/28/2026 04:41:37 - INFO - qdiff.optimization.model_recon -   Block 11 optimization complete
+01/28/2026 04:41:37 - INFO - qdiff.optimization.model_recon - Processing WAN block 12/39
+01/28/2026 04:41:37 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 04:41:37 - INF01/28/2026 04:41:45 - INFO - qdiff.optimization.model_recon -   Block input shape01/28/2026 04:42:39 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 04:42:39 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 04:42:40 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.79GB
+01/28/2026 04:42:01/28/2026 04:43:42 - INFO - qdiff.optimization.model_recon -   Block 12 iter 50/01/28/2026 04:44:301/28/2026 04:45:38 - INFO - qdiff.optimization.model_recon -   Block 12 iter 10001/28/2026 04:46:3501/28/2026 04:47:32 - INFO - qdiff.optimization.model_recon -   Block 12 iter 15001/28/2026 04:48:2701/28/2026 04:49:26 - INFO - qdiff.optimization.model_recon -   Block 12 iter 20001/28/2026 04:50:2201/28/2026 04:51:19 - INFO - qdiff.optimization.model_recon -   Block 12 iter 25001/28/2026 04:52:1301/28/2026 04:53:11 - INFO - qdiff.optimization.model_recon -   Block 12 iter 30001/28/2026 04:54:0501/28/2026 04:55:01 - INFO - qdiff.optimization.model_recon -   Block 12 iter 35001/28/2026 04:55:5801/28/2026 04:56:54 - INFO - qdiff.optimization.model_recon -   Block 12 iter 40001/28/2026 04:57:5101/28/2026 04:58:46 - INFO - qdiff.optimization.model_recon -   Block 12 iter 45001/28/2026 04:59:4501/28/2026 05:00:38 - INFO - qdiff.optimization.model_recon -   Block 12 iter 50001/28/2026 05:01:4001/28/2026 05:02:33 - INFO - qdiff.optimization.model_recon -   Block 12 iter 55001/28/2026 05:03:3601/28/2026 05:04:27 - INFO - qdiff.optimization.model_recon -   Block 12 iter 60001/28/2026 05:05:3501/28/2026 05:06:23 - INFO - qdiff.optimization.model_recon -   Block 12 iter 65001/28/2026 05:07:3001/28/2026 05:08:17 - INFO - qdiff.optimization.model_recon -   Block 12 iter 70001/28/2026 05:09:2401/28/2026 05:10:10 - INFO - qdiff.optimization.model_recon -   Block 12 iter 75001/28/2026 05:11:1601/28/2026 05:12:02 - INFO - qdiff.optimization.model_recon -   Block 12 iter 80001/28/2026 05:13:0701/28/2026 05:13:53 - INFO - qdiff.optimization.model_recon -   Block 12 iter 85001/28/2026 05:14:5901/28/2026 05:15:44 - INFO - qdiff.optimization.model_recon -   Block 12 iter 90001/28/2026 05:16:5201/28/2026 05:17:37 - INFO - qdiff.optimization.model_recon -   Block 12 iter 95001/28/2026 05:18:5001/28/2026 05:19:30 - INFO - qdiff.optimization.model_recon -   Block 12 optimiza01/28/2026 05:20:46 - INFO - qdiff.optimization.model_recon -   Block 12 optimization complete
+01/28/2026 05:20:46 - INFO - qdiff.optimization.model_recon - Processing WAN block 13/39
+01/28/2026 05:20:46 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 05:20:46 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 13...
+01/28/2026 05:21:53 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 05:21:53 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 05:21:53 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.80GB
+01/28/2026 05:21:01/28/2026 05:22:38 - INFO - qdiff.optimization.model_recon -   Block 13 iter 50/01/28/2026 05:23:501/28/2026 05:24:36 - INFO - qdiff.optimization.model_recon -   Block 13 iter 10001/28/2026 05:25:5201/28/2026 05:26:33 - INFO - qdiff.optimization.model_recon -   Block 13 iter 15001/28/2026 05:27:4701/28/2026 05:28:30 - INFO - qdiff.optimization.model_recon -   Block 13 iter 20001/28/2026 05:29:4101/28/2026 05:30:22 - INFO - qdiff.optimization.model_recon -   Block 13 iter 25001/28/2026 05:31:3201/28/2026 05:32:12 - INFO - qdiff.optimization.model_recon -   Block 13 iter 30001/28/2026 05:33:2301/28/2026 05:34:06 - INFO - qdiff.optimization.model_recon -   Block 13 iter 35001/28/2026 05:35:1401/28/2026 05:35:56 - INFO - qdiff.optimization.model_recon -   Block 13 iter 40001/28/2026 05:37:0601/28/2026 05:37:48 - INFO - qdiff.optimization.model_recon -   Block 13 iter 45001/28/2026 05:38:5901/28/2026 05:39:41 - INFO - qdiff.optimization.model_recon -   Block 13 iter 50001/28/2026 05:40:5201/28/2026 05:41:36 - INFO - qdiff.optimization.model_recon -   Block 13 iter 55001/28/2026 05:42:4701/28/2026 05:43:29 - INFO - qdiff.optimization.model_recon -   Block 13 iter 60001/28/2026 05:44:4101/28/2026 05:45:23 - INFO - qdiff.optimization.model_recon -   Block 13 iter 65001/28/2026 05:46:3901/28/2026 05:47:23 - INFO - qdiff.optimization.model_recon -   Block 13 iter 70001/28/2026 05:48:3501/28/2026 05:49:17 - INFO - qdiff.optimization.model_recon -   Block 13 iter 75001/28/2026 05:50:2801/28/2026 05:51:11 - INFO - qdiff.optimization.model_recon -   Block 13 iter 80001/28/2026 05:52:2001/28/2026 05:53:01 - INFO - qdiff.optimization.model_recon -   Block 13 iter 85001/28/2026 05:54:1101/28/2026 05:54:52 - INFO - qdiff.optimization.model_recon -   Block 13 iter 90001/28/2026 05:56:0201/28/2026 05:56:42 - INFO - qdiff.optimization.model_recon -   Block 13 iter 95001/28/2026 05:57:5501/28/2026 05:58:35 - INFO - qdiff.optimization.model_recon -   Block 13 optimiza01/28/2026 05:59:48 - INFO - qdiff.optimization.model_recon -   Block 13 optimization complete
+01/28/2026 05:59:48 - INFO - qdiff.optimization.model_recon - Processing WAN block 14/39
+01/28/2026 05:59:48 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 05:59:48 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 14...
+01/28/2026 06:00:58 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 06:00:58 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 06:00:59 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.82GB
+01/28/2026 06:01:01/28/2026 06:01:40 - INFO - qdiff.optimization.model_recon -   Block 14 iter 50/01/28/2026 06:02:501/28/2026 06:03:38 - INFO - qdiff.optimization.model_recon -   Block 14 iter 10001/28/2026 06:04:5201/28/2026 06:05:32 - INFO - qdiff.optimization.model_recon -   Block 14 iter 15001/28/2026 06:06:4601/28/2026 06:07:25 - INFO - qdiff.optimization.model_recon -   Block 14 iter 20001/28/2026 06:08:4101/28/2026 06:09:18 - INFO - qdiff.optimization.model_recon -   Block 14 iter 25001/28/2026 06:10:3301/28/2026 06:11:12 - INFO - qdiff.optimization.model_recon -   Block 14 iter 30001/28/2026 06:12:2601/28/2026 06:13:02 - INFO - qdiff.optimization.model_recon -   Block 14 iter 35001/28/2026 06:14:1601/28/2026 06:14:52 - INFO - qdiff.optimization.model_recon -   Block 14 iter 40001/28/2026 06:16:0801/28/2026 06:16:41 - INFO - qdiff.optimization.model_recon -   Block 14 iter 45001/28/2026 06:18:0001/28/2026 06:18:34 - INFO - qdiff.optimization.model_recon -   Block 14 iter 50001/28/2026 06:19:5201/28/2026 06:20:28 - INFO - qdiff.optimization.model_recon -   Block 14 iter 55001/28/2026 06:21:5001/28/2026 06:22:28 - INFO - qdiff.optimization.model_recon -   Block 14 iter 60001/28/2026 06:23:4801/28/2026 06:24:24 - INFO - qdiff.optimization.model_recon -   Block 14 iter 65001/28/2026 06:25:4201/28/2026 06:26:19 - INFO - qdiff.optimization.model_recon -   Block 14 iter 70001/28/2026 06:27:3601/28/2026 06:28:13 - INFO - qdiff.optimization.model_recon -   Block 14 iter 75001/28/2026 06:29:2901/28/2026 06:30:05 - INFO - qdiff.optimization.model_recon -   Block 14 iter 80001/28/2026 06:31:2001/28/2026 06:31:56 - INFO - qdiff.optimization.model_recon -   Block 14 iter 85001/28/2026 06:33:1201/28/2026 06:33:48 - INFO - qdiff.optimization.model_recon -   Block 14 iter 90001/28/2026 06:35:0301/28/2026 06:35:38 - INFO - qdiff.optimization.model_recon -   Block 14 iter 95001/28/2026 06:36:5401/28/2026 06:37:27 - INFO - qdiff.optimization.model_recon -   Block 14 optimiza01/28/2026 06:38:44 - INFO - qdiff.optimization.model_recon -   Block 14 optimization complete
+01/28/2026 06:38:44 - INFO - qdiff.optimization.model_recon - Processing WAN block 15/39
+01/28/2026 06:38:44 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 06:38:44 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 15...
+01/28/2026 06:39:58 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 06:39:58 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 06:39:58 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.83GB
+01/28/2026 06:40:01/28/2026 06:40:36 - INFO - qdiff.optimization.model_recon -   Block 15 iter 50/01/28/2026 06:41:501/28/2026 06:42:31 - INFO - qdiff.optimization.model_recon -   Block 15 iter 10001/28/2026 06:43:4801/28/2026 06:44:26 - INFO - qdiff.optimization.model_recon -   Block 15 iter 15001/28/2026 06:45:4201/28/2026 06:46:22 - INFO - qdiff.optimization.model_recon -   Block 15 iter 20001/28/2026 06:47:3601/28/2026 06:48:16 - INFO - qdiff.optimization.model_recon -   Block 15 iter 25001/28/2026 06:49:2901/28/2026 06:50:11 - INFO - qdiff.optimization.model_recon -   Block 15 iter 30001/28/2026 06:51:2201/28/2026 06:52:03 - INFO - qdiff.optimization.model_recon -   Block 15 iter 35001/28/2026 06:53:1301/28/2026 06:53:54 - INFO - qdiff.optimization.model_recon -   Block 15 iter 40001/28/2026 06:55:0501/28/2026 06:55:44 - INFO - qdiff.optimization.model_recon -   Block 15 iter 45001/28/2026 06:56:5601/28/2026 06:57:35 - INFO - qdiff.optimization.model_recon -   Block 15 iter 50001/28/2026 06:58:4801/28/2026 06:59:27 - INFO - qdiff.optimization.model_recon -   Block 15 iter 55001/28/2026 07:00:4101/28/2026 07:01:20 - INFO - qdiff.optimization.model_recon -   Block 15 iter 60001/28/2026 07:02:3601/28/2026 07:03:15 - INFO - qdiff.optimization.model_recon -   Block 15 iter 65001/28/2026 07:04:2901/28/2026 07:05:10 - INFO - qdiff.optimization.model_recon -   Block 15 iter 70001/28/2026 07:06:2301/28/2026 07:07:04 - INFO - qdiff.optimization.model_recon -   Block 15 iter 75001/28/2026 07:08:2001/28/2026 07:09:02 - INFO - qdiff.optimization.model_recon -   Block 15 iter 80001/28/2026 07:10:1401/28/2026 07:10:58 - INFO - qdiff.optimization.model_recon -   Block 15 iter 85001/28/2026 07:12:1001/28/2026 07:12:53 - INFO - qdiff.optimization.model_recon -   Block 15 iter 90001/28/2026 07:14:0101/28/2026 07:14:43 - INFO - qdiff.optimization.model_recon -   Block 15 iter 95001/28/2026 07:15:5101/28/2026 07:16:35 - INFO - qdiff.optimization.model_recon -   Block 15 optimiza01/28/2026 07:17:42 - INFO - qdiff.optimization.model_recon -   Block 15 optimization complete
+01/28/2026 07:17:42 - INFO - qdiff.optimization.model_recon - Processing WAN block 16/39
+01/28/2026 07:17:42 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 07:17:42 - INF01/28/2026 07:17:53 - INFO - qdiff.optimization.model_recon -   Block input shape01/28/2026 07:19:00 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 07:19:00 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 07:19:01 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.85GB
+01/28/2026 07:19:01/28/2026 07:19:47 - INFO - qdiff.optimization.model_recon -   Block 16 iter 50/01/28/2026 07:20:501/28/2026 07:21:38 - INFO - qdiff.optimization.model_recon -   Block 16 iter 10001/28/2026 07:22:4601/28/2026 07:23:30 - INFO - qdiff.optimization.model_recon -   Block 16 iter 15001/28/2026 07:24:3901/28/2026 07:25:23 - INFO - qdiff.optimization.model_recon -   Block 16 iter 20001/28/2026 07:26:3201/28/2026 07:27:16 - INFO - qdiff.optimization.model_recon -   Block 16 iter 25001/28/2026 07:28:2601/28/2026 07:29:11 - INFO - qdiff.optimization.model_recon -   Block 16 iter 30001/28/2026 07:30:2001/28/2026 07:31:06 - INFO - qdiff.optimization.model_recon -   Block 16 iter 35001/28/2026 07:32:1301/28/2026 07:33:00 - INFO - qdiff.optimization.model_recon -   Block 16 iter 40001/28/2026 07:34:0501/28/2026 07:34:51 - INFO - qdiff.optimization.model_recon -   Block 16 iter 45001/28/2026 07:35:5501/28/2026 07:36:42 - INFO - qdiff.optimization.model_recon -   Block 16 iter 50001/28/2026 07:37:4601/28/2026 07:38:36 - INFO - qdiff.optimization.model_recon -   Block 16 iter 55001/28/2026 07:39:3901/28/2026 07:40:27 - INFO - qdiff.optimization.model_recon -   Block 16 iter 60001/28/2026 07:41:3101/28/2026 07:42:19 - INFO - qdiff.optimization.model_recon -   Block 16 iter 65001/28/2026 07:43:2701/28/2026 07:44:14 - INFO - qdiff.optimization.model_recon -   Block 16 iter 70001/28/2026 07:45:2401/28/2026 07:46:09 - INFO - qdiff.optimization.model_recon -   Block 16 iter 75001/28/2026 07:47:2001/28/2026 07:48:06 - INFO - qdiff.optimization.model_recon -   Block 16 iter 80001/28/2026 07:49:1601/28/2026 07:50:04 - INFO - qdiff.optimization.model_recon -   Block 16 iter 85001/28/2026 07:51:1301/28/2026 07:51:59 - INFO - qdiff.optimization.model_recon -   Block 16 iter 90001/28/2026 07:53:0501/28/2026 07:53:51 - INFO - qdiff.optimization.model_recon -   Block 16 iter 95001/28/2026 07:54:5701/28/2026 07:55:41 - INFO - qdiff.optimization.model_recon -   Block 16 optimiza01/28/2026 07:56:47 - INFO - qdiff.optimization.model_recon -   Block 16 optimization complete
+01/28/2026 07:56:47 - INFO - qdiff.optimization.model_recon - Processing WAN block 17/39
+01/28/2026 07:56:47 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 07:56:47 - INF01/28/2026 07:57:03 - INFO - qdiff.optimization.model_recon -   Block input shape01/28/2026 07:58:09 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 07:58:09 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 07:58:10 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.86GB
+01/28/2026 07:58:01/28/2026 07:58:59 - INFO - qdiff.optimization.model_recon -   Block 17 iter 50/01/28/2026 08:00:001/28/2026 08:00:53 - INFO - qdiff.optimization.model_recon -   Block 17 iter 10001/28/2026 08:01:5901/28/2026 08:02:46 - INFO - qdiff.optimization.model_recon -   Block 17 iter 15001/28/2026 08:03:5301/28/2026 08:04:41 - INFO - qdiff.optimization.model_recon -   Block 17 iter 20001/28/2026 08:05:4701/28/2026 08:06:39 - INFO - qdiff.optimization.model_recon -   Block 17 iter 25001/28/2026 08:07:4401/28/2026 08:08:35 - INFO - qdiff.optimization.model_recon -   Block 17 iter 30001/28/2026 08:09:4001/28/2026 08:10:34 - INFO - qdiff.optimization.model_recon -   Block 17 iter 35001/28/2026 08:11:3901/28/2026 08:12:34 - INFO - qdiff.optimization.model_recon -   Block 17 iter 40001/28/2026 08:13:3401/28/2026 08:14:29 - INFO - qdiff.optimization.model_recon -   Block 17 iter 45001/28/2026 08:15:3101/28/2026 08:16:29 - INFO - qdiff.optimization.model_recon -   Block 17 iter 50001/28/2026 08:17:2801/28/2026 08:18:25 - INFO - qdiff.optimization.model_recon -   Block 17 iter 55001/28/2026 08:19:2301/28/2026 08:20:17 - INFO - qdiff.optimization.model_recon -   Block 17 iter 60001/28/2026 08:21:1501/28/2026 08:22:09 - INFO - qdiff.optimization.model_recon -   Block 17 iter 65001/28/2026 08:23:1001/28/2026 08:24:05 - INFO - qdiff.optimization.model_recon -   Block 17 iter 70001/28/2026 08:25:0401/28/2026 08:25:58 - INFO - qdiff.optimization.model_recon -   Block 17 iter 75001/28/2026 08:26:5701/28/2026 08:27:52 - INFO - qdiff.optimization.model_recon -   Block 17 iter 80001/28/2026 08:28:5101/28/2026 08:29:49 - INFO - qdiff.optimization.model_recon -   Block 17 iter 85001/28/2026 08:30:4501/28/2026 08:31:41 - INFO - qdiff.optimization.model_recon -   Block 17 iter 90001/28/2026 08:32:3801/28/2026 08:33:34 - INFO - qdiff.optimization.model_recon -   Block 17 iter 95001/28/2026 08:34:3101/28/2026 08:35:26 - INFO - qdiff.optimization.model_recon -   Block 17 optimiza01/28/2026 08:36:23 - INFO - qdiff.optimization.model_recon -   Block 17 optimization complete
+01/28/2026 08:36:23 - INFO - qdiff.optimization.model_recon - Processing WAN block 18/39
+01/28/2026 08:36:23 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 08:36:23 - INF01/28/2026 08:36:53 - INFO - qdiff.optimization.model_recon -   Block input shape01/28/2026 08:37:50 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 08:37:50 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 08:37:50 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.88GB
+01/28/2026 08:37:01/28/2026 08:38:52 - INFO - qdiff.optimization.model_recon -   Block 18 iter 50/01/28/2026 08:39:401/28/2026 08:40:45 - INFO - qdiff.optimization.model_recon -   Block 18 iter 10001/28/2026 08:41:4001/28/2026 08:42:40 - INFO - qdiff.optimization.model_recon -   Block 18 iter 15001/28/2026 08:43:3301/28/2026 08:44:33 - INFO - qdiff.optimization.model_recon -   Block 18 iter 20001/28/2026 08:45:2601/28/2026 08:46:26 - INFO - qdiff.optimization.model_recon -   Block 18 iter 25001/28/2026 08:47:1901/28/2026 08:48:20 - INFO - qdiff.optimization.model_recon -   Block 18 iter 30001/28/2026 08:49:1201/28/2026 08:50:14 - INFO - qdiff.optimization.model_recon -   Block 18 iter 35001/28/2026 08:51:0401/28/2026 08:52:07 - INFO - qdiff.optimization.model_recon -   Block 18 iter 40001/28/2026 08:53:0101/28/2026 08:54:04 - INFO - qdiff.optimization.model_recon -   Block 18 iter 45001/28/2026 08:54:5401/28/2026 08:55:59 - INFO - qdiff.optimization.model_recon -   Block 18 iter 50001/28/2026 08:56:4801/28/2026 08:57:57 - INFO - qdiff.optimization.model_recon -   Block 18 iter 55001/28/2026 08:58:4301/28/2026 08:59:53 - INFO - qdiff.optimization.model_recon -   Block 18 iter 60001/28/2026 09:00:4001/28/2026 09:01:51 - INFO - qdiff.optimization.model_recon -   Block 18 iter 65001/28/2026 09:02:3601/28/2026 09:03:48 - INFO - qdiff.optimization.model_recon -   Block 18 iter 70001/28/2026 09:04:3101/28/2026 09:05:40 - INFO - qdiff.optimization.model_recon -   Block 18 iter 75001/28/2026 09:06:2301/28/2026 09:07:33 - INFO - qdiff.optimization.model_recon -   Block 18 iter 80001/28/2026 09:08:1601/28/2026 09:09:25 - INFO - qdiff.optimization.model_recon -   Block 18 iter 85001/28/2026 09:10:0801/28/2026 09:11:17 - INFO - qdiff.optimization.model_recon -   Block 18 iter 90001/28/2026 09:12:0001/28/2026 09:13:11 - INFO - qdiff.optimization.model_recon -   Block 18 iter 95001/28/2026 09:13:5401/28/2026 09:15:01 - INFO - qdiff.optimization.model_recon -   Block 18 optimiza01/28/2026 09:15:45 - INFO - qdiff.optimization.model_recon -   Block 18 optimization complete
+01/28/2026 09:15:45 - INFO - qdiff.optimization.model_recon - Processing WAN block 19/39
+01/28/2026 09:15:45 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 09:15:45 - INF01/28/2026 09:16:31 - INFO - qdiff.optimization.model_recon -   Block input shape01/28/2026 09:17:15 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 09:17:15 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 09:17:15 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.89GB
+01/28/2026 09:17:01/28/2026 09:18:29 - INFO - qdiff.optimization.model_recon -   Block 19 iter 50/01/28/2026 09:19:101/28/2026 09:20:22 - INFO - qdiff.optimization.model_recon -   Block 19 iter 10001/28/2026 09:21:0301/28/2026 09:22:14 - INFO - qdiff.optimization.model_recon -   Block 19 iter 15001/28/2026 09:22:5501/28/2026 09:24:06 - INFO - qdiff.optimization.model_recon -   Block 19 iter 20001/28/2026 09:24:4901/28/2026 09:26:01 - INFO - qdiff.optimization.model_recon -   Block 19 iter 25001/28/2026 09:26:4201/28/2026 09:27:51 - INFO - qdiff.optimization.model_recon -   Block 19 iter 30001/28/2026 09:28:3401/28/2026 09:29:43 - INFO - qdiff.optimization.model_recon -   Block 19 iter 35001/28/2026 09:30:2601/28/2026 09:31:35 - INFO - qdiff.optimization.model_recon -   Block 19 iter 40001/28/2026 09:32:1801/28/2026 09:33:29 - INFO - qdiff.optimization.model_recon -   Block 19 iter 45001/28/2026 09:34:1301/28/2026 09:35:23 - INFO - qdiff.optimization.model_recon -   Block 19 iter 50001/28/2026 09:36:0701/28/2026 09:37:17 - INFO - qdiff.optimization.model_recon -   Block 19 iter 55001/28/2026 09:38:0001/28/2026 09:39:10 - INFO - qdiff.optimization.model_recon -   Block 19 iter 60001/28/2026 09:39:5301/28/2026 09:41:04 - INFO - qdiff.optimization.model_recon -   Block 19 iter 65001/28/2026 09:41:5001/28/2026 09:42:59 - INFO - qdiff.optimization.model_recon -   Block 19 iter 70001/28/2026 09:43:4501/28/2026 09:44:54 - INFO - qdiff.optimization.model_recon -   Block 19 iter 75001/28/2026 09:45:3901/28/2026 09:46:49 - INFO - qdiff.optimization.model_recon -   Block 19 iter 80001/28/2026 09:47:3201/28/2026 09:48:43 - INFO - qdiff.optimization.model_recon -   Block 19 iter 85001/28/2026 09:49:2401/28/2026 09:50:35 - INFO - qdiff.optimization.model_recon -   Block 19 iter 90001/28/2026 09:51:1501/28/2026 09:52:29 - INFO - qdiff.optimization.model_recon -   Block 19 iter 95001/28/2026 09:53:1101/28/2026 09:54:19 - INFO - qdiff.optimization.model_recon -   Block 19 optimiza01/28/2026 09:55:02 - INFO - qdiff.optimization.model_recon -   Block 19 optimization complete
+01/28/2026 09:55:02 - INFO - qdiff.optimization.model_recon - Completed 20/40 blocks
+01/28/2026 09:55:02 - INFO - qdiff.optimization.model_recon - Processing WAN block 20/39
+01/28/2026 09:55:02 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 09:55:02 - INF01/28/2026 09:55:52 - INFO - qdiff.optimization.model_recon -   Block input shape01/28/2026 09:56:36 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 09:56:36 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 09:56:36 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.91GB
+01/28/2026 09:56:01/28/2026 09:57:46 - INFO - qdiff.optimization.model_recon -   Block 20 iter 50/01/28/2026 09:58:301/28/2026 09:59:37 - INFO - qdiff.optimization.model_recon -   Block 20 iter 10001/28/2026 10:00:2201/28/2026 10:01:29 - INFO - qdiff.optimization.model_recon -   Block 20 iter 15001/28/2026 10:02:1501/28/2026 10:03:21 - INFO - qdiff.optimization.model_recon -   Block 20 iter 20001/28/2026 10:04:0601/28/2026 10:05:13 - INFO - qdiff.optimization.model_recon -   Block 20 iter 25001/28/2026 10:05:5801/28/2026 10:07:04 - INFO - qdiff.optimization.model_recon -   Block 20 iter 30001/28/2026 10:07:5001/28/2026 10:08:55 - INFO - qdiff.optimization.model_recon -   Block 20 iter 35001/28/2026 10:09:4201/28/2026 10:10:47 - INFO - qdiff.optimization.model_recon -   Block 20 iter 40001/28/2026 10:11:3501/28/2026 10:12:38 - INFO - qdiff.optimization.model_recon -   Block 20 iter 45001/28/2026 10:13:2801/28/2026 10:14:30 - INFO - qdiff.optimization.model_recon -   Block 20 iter 50001/28/2026 10:15:2101/28/2026 10:16:21 - INFO - qdiff.optimization.model_recon -   Block 20 iter 55001/28/2026 10:17:1301/28/2026 10:18:13 - INFO - qdiff.optimization.model_recon -   Block 20 iter 60001/28/2026 10:19:0501/28/2026 10:20:05 - INFO - qdiff.optimization.model_recon -   Block 20 iter 65001/28/2026 10:20:5701/28/2026 10:21:57 - INFO - qdiff.optimization.model_recon -   Block 20 iter 70001/28/2026 10:22:4901/28/2026 10:23:49 - INFO - qdiff.optimization.model_recon -   Block 20 iter 75001/28/2026 10:24:4101/28/2026 10:25:41 - INFO - qdiff.optimization.model_recon -   Block 20 iter 80001/28/2026 10:26:3301/28/2026 10:27:34 - INFO n.model_recon -   Block 20 iter 800/1000: GPU 27.95GB
+01/28/2026 10:28:26 - INFO - qdiff.optimization.model_recon -   Block 20 iter 850/1000: GPU 27.95GB
+01/28/2026 10:30:19 - INFO - qdiff.optimization.model_recon -   Block 20 iter 900/1000: GPU 27.95GB
+01/28/2026 10:32:13 - INFO - qdiff.optimization.model_recon -   Block 20 iter 950/1000: GPU 27.95GB
+01/28/2026 10:34:05 - INFO - qdiff.optimization.model_recon -   Block 20 optimization complete
+01/28/2026 10:34:05 - INFO - qdiff.optimization.model_recon - Processing WAN block 21/39
+01/28/2026 10:34:05 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 10:34:05 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 21...
+01/28/2026 10:35:42 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 10:35:42 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 10:35:43 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.92GB
+01/28/2026 10:35:45 - INFO - qdiff.optimization.model_recon -   Block 21 iter 0/1000: GPU 27.97GB
+01/28/2026 10:37:37 - INFO - qdiff.optimization.model_recon -   Block 21 iter 50/1000: GPU 27.97GB
+01/28/2026 10:39:28 - INFO - qdiff.optimization.model_recon -   Block 21 iter 100/1000: GPU 27.97GB
+01/28/2026 10:41:19 - INFO - qdiff.optimization.model_recon -   Block 21 iter 150/1000: GPU 27.97GB
+01/28/2026 10:43:11 - INFO - qdiff.optimization.model_recon -   Block 21 iter 200/1000: GPU 27.97GB
+01/28/2026 10:45:02 - INFO - qdiff.optimization.model_recon -   Block 21 iter 250/1000: GPU 27.97GB
+01/28/2026 10:46:54 - INFO - qdiff.optimization.model_recon -   Block 21 iter 300/1000: GPU 27.97GB
+01/28/2026 10:48:46 - INFO - qdiff.optimization.model_recon -   Block 21 iter 350/1000: GPU 27.97GB
+01/28/2026 10:50:39 - INFO - qdiff.optimization.model_recon -   Block 21 iter 400/1000: GPU 27.97GB
+01/28/2026 10:52:30 - INFO - qdiff.optimization.model_recon -   Block 21 iter 450/1000: GPU 27.97GB
+01/28/2026 10:54:23 - INFO - qdiff.optimization.model_recon -   Block 21 iter 500/1000: GPU 27.97GB
+01/28/2026 10:56:15 - INFO - qdiff.optimization.model_recon -   Block 21 iter 550/1000: GPU 27.97GB
+01/28/2026 10:58:07 - INFO - qdiff.optimization.model_recon -   Block 21 iter 600/1000: GPU 27.97GB
+01/28/2026 11:00:00 - INFO - qdiff.optimization.model_recon -   Block 21 iter 650/1000: GPU 27.97GB
+01/28/2026 11:01:52 - INFO - qdiff.optimization.model_recon -   Block 21 iter 700/1000: GPU 27.97GB
+01/28/2026 11:03:44 - INFO - qdiff.optimization.model_recon -   Block 21 iter 750/1000: GPU 27.97GB
+01/28/2026 11:05:37 - INFO - qdiff.optimization.model_recon -   Block 21 iter 800/1000: GPU 27.97GB
+01/28/2026 11:07:29 - INFO - qdiff.optimization.model_recon -   Block 21 iter 850/1000: GPU 27.97GB
+01/28/2026 11:09:20 - INFO - qdiff.optimization.model_recon -   Block 21 iter 900/1000: GPU 27.97GB
+01/28/2026 11:11:12 - INFO - qdiff.optimization.model_recon -   Block 21 iter 950/1000: GPU 27.97GB
+01/28/2026 11:13:03 - INFO - qdiff.optimization.model_recon -   Block 21 optimization complete
+01/28/2026 11:13:03 - INFO - qdiff.optimization.model_recon - Processing WAN block 22/39
+01/28/2026 11:13:03 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 11:13:03 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 22...
+01/28/2026 11:14:45 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 11:14:45 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 11:14:46 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 27.94GB
+01/28/2026 11:14:48 - INFO - qdiff.optimization.model_recon -   Block 22 iter 0/1000: GPU 27.98GB
+01/28/2026 11:16:39 - INFO - qdiff.optimization.model_recon -   Block 22 iter 50/1000: GPU 27.98GB
+01/28/2026 15:17:49 - INFO - qdiff.optimization.model_recon -   Block 28 iter 200/1000: GPU 28.07GB
+01/28/2026 15:19:41 - INFO - qdiff.optimization.model_recon -   Block 28 iter 250/1000: GPU 28.07GB
+01/28/2026 15:21:33 - INFO - qdiff.optimization.model_recon -   Block 28 iter 300/1000: GPU 28.07GB
+01/28/2026 15:23:24 - INFO - qdiff.optimization.model_recon -   Block 28 iter 350/1000: GPU 28.07GB
+01/28/2026 15:25:16 - INFO - qdiff.optimization.model_recon -   Block 28 iter 400/1000: GPU 28.07GB
+01/28/2026 15:27:08 - INFO - qdiff.optimization.model_recon -   Block 28 iter 450/1000: GPU 28.07GB
+01/28/2026 15:29:00 - INFO - qdiff.optimization.model_recon -   Block 28 iter 500/1000: GPU 28.07GB
+01/28/2026 15:30:52 - INFO - qdiff.optimization.model_recon -   Block 28 iter 550/1000: GPU 28.07GB
+01/28/2026 15:32:45 - INFO - qdiff.optimization.model_recon -   Block 28 iter 600/1000: GPU 28.07GB
+01/28/2026 15:34:38 - INFO - qdiff.optimization.model_recon -   Block 28 iter 650/1000: GPU 28.07GB
+01/28/2026 15:36:30 - INFO - qdiff.optimization.model_recon -   Block 28 iter 700/1000: GPU 28.07GB
+01/28/2026 15:38:22 - INFO - qdiff.optimization.model_recon -   Block 28 iter 750/1000: GPU 28.07GB
+01/28/2026 15:40:15 - INFO - qdiff.optimization.model_recon -   Block 28 iter 800/1000: GPU 28.07GB
+01/28/2026 15:42:08 - INFO - qdiff.optimization.model_recon -   Block 28 iter 850/1000: GPU 28.07GB
+01/28/2026 15:44:02 - INFO - qdiff.optimization.model_recon -   Block 28 iter 900/1000: GPU 28.07GB
+01/28/2026 15:45:56 - INFO - qdiff.optimization.model_recon -   Block 28 iter 950/1000: GPU 28.07GB
+01/28/2026 15:47:47 - INFO - qdiff.optimization.model_recon -   Block 28 optimization complete
+01/28/2026 15:47:48 - INFO - qdiff.optimization.model_recon - Processing WAN block 29/39
+01/28/2026 15:47:48 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 15:47:48 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 29...
+01/28/2026 15:49:57 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 15:49:57 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 15:49:57 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.04GB
+01/28/2026 15:50:00 - INFO - qdiff.optimization.model_recon -   Block 29 iter 0/1000: GPU 28.08GB
+01/28/2026 15:51:55 - INFO - qdiff.optimization.model_recon -   Block 29 iter 50/1000: GPU 28.08GB
+01/28/2026 15:53:49 - INFO - qdiff.optimization.model_recon -   Block 29 iter 100/1000: GPU 28.08GB
+01/28/2026 15:55:41 - INFO - qdiff.optimization.model_recon -   Block 29 iter 150/1000: GPU 28.08GB
+01/28/2026 15:57:33 - INFO - qdiff.optimization.model_recon -   Block 29 iter 200/1000: GPU 28.08GB
+01/28/2026 15:59:26 - INFO - qdiff.optimization.model_recon -   Block 29 iter 250/1000: GPU 28.08GB
+01/28/2026 16:01:18 - INFO - qdiff.optimization.model_recon -   Block 29 iter 300/1000: GPU 28.08GB
+01/28/2026 16:03:10 - INFO - qdiff.optimization.model_recon -   Block 29 iter 350/1000: GPU 28.08GB
+01/28/2026 16:05:03 - INFO - qdiff.optimization.model_recon -   Block 29 iter 400/1000: GPU 28.08GB
+01/28/2026 16:06:56 - INFO - qdiff.optimization.model_recon -   Block 29 iter 450/1000: GPU 28.08GB
+01/28/2026 16:08:48 - INFO - qdiff.optimization.model_recon -   Block 29 iter 500/1000: GPU 28.08GB
+01/28/2026 16:10:41 - INFO - qdiff.optimization.model_recon -   Block 29 iter 550/1000: GPU 28.08GB
+01/28/2026 16:12:32 - INFO - qdiff.optimization.model_recon -   Block 29 iter 600/1000: GPU 28.08GB
+01/28/2026 16:14:24 - INFO - qdiff.optimization.model_recon -   Block 29 iter 650/1000: GPU 28.08GB
+01/28/2026 16:16:16 - INFO - qdiff.optimization.model_recon -   Block 29 iter 700/1000: GPU 28.08GB
+01/28/2026 16:18:09 - INFO - qdiff.optimization.model_recon -   Block 29 iter 750/1000: GPU 28.08GB
+01/28/2026 16:20:01 - INFO - qdiff.optimization.model_recon -   Block 29 iter 800/1000: GPU 28.08GB
+01/28/2026 16:21:53 - INFO - qdiff.optimization.model_recon -   Block 29 iter 850/1000: GPU 28.08GB
+01/28/2026 16:23:45 - INFO - qdiff.optimization.model_recon -   Block 29 iter 900/1000: GPU 28.08GB
+01/28/2026 16:25:38 - INFO - qdiff.optimization.model_recon -   Block 29 iter 950/1000: GPU 28.08GB
+01/28/2026 16:27:30 - INFO - qdiff.optimization.model_recon -   Block 29 optimization complete
+01/28/2026 16:27:30 - INFO - qdiff.optimization.model_recon - Completed 30/40 blocks
+01/28/2026 16:27:30 - INFO - qdiff.optimization.model_recon - Processing WAN block 30/39
+01/28/2026 16:27:30 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 16:27:30 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 30...
+01/28/2026 16:29:45 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 16:29:45 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 16:29:45 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.05GB
+01/28/2026 16:29:47 - INFO - qdiff.optimization.model_recon -   Block 30 iter 0/1000: GPU 28.10GB
+01/28/2026 16:31:40 - INFO - qdiff.optimization.model_recon -   Block 30 iter 50/1000: GPU 28.10GB
+01/28/2026 16:33:32 - INFO - qdiff.optimization.model_recon -   Block 30 iter 100/1000: GPU 28.10GB
+01/28/2026 16:35:27 - INFO - qdiff.optimization.model_recon -   Block 30 iter 150/1000: GPU 28.10GB
+01/28/2026 16:37:20 - INFO - qdiff.optimization.model_recon -   Block 30 iter 200/1000: GPU 28.10GB
+01/28/2026 16:39:13 - INFO - qdiff.optimization.model_recon -   Block 30 iter 250/1000: GPU 28.10GB
+01/28/2026 16:41:05 - INFO - qdiff.optimization.model_recon -   Block 30 iter 300/1000: GPU 28.10GB
+01/28/2026 16:42:59 - INFO - qdiff.optimization.model_recon -   Block 30 iter 350/1000: GPU 28.10GB
+01/28/2026 16:44:53 - INFO - qdiff.optimization.model_recon -   Block 30 iter 400/1000: GPU 28.10GB
+01/28/2026 16:46:47 - INFO - qdiff.optimization.model_recon -   Block 30 iter 450/1000: GPU 28.10GB
+01/28/2026 16:48:41 - INFO - qdiff.optimization.model_recon -   Block 30 iter 500/1000: GPU 28.10GB
+01/28/2026 16:50:34 - INFO - qdiff.optimization.model_recon -   Block 30 iter 550/1000: GPU 28.10GB
+01/28/2026 16:52:26 - INFO - qdiff.optimization.model_recon -   Block 30 iter 600/1000: GPU 28.10GB
+01/28/2026 16:54:18 - INFO - qdiff.optimization.model_recon -   Block 30 iter 650/1000: GPU 28.10GB
+01/28/2026 16:56:12 - INFO - qdiff.optimization.model_recon -   Block 30 iter 700/1000: GPU 28.10GB
+01/28/2026 16:58:05 - INFO - qdiff.optimization.model_recon -   Block 30 iter 750/1000: GPU 28.10GB
+01/28/2026 16:59:59 - INFO - qdiff.optimization.model_recon -   Block 30 iter 800/1000: GPU 28.10GB
+01/28/2026 17:01:52 - INFO - qdiff.optimization.model_recon -   Block 30 iter 850/1000: GPU 28.10GB
+01/28/2026 17:03:45 - INFO - qdiff.optimization.model_recon -   Block 30 iter 900/1000: GPU 28.10GB
+01/28/2026 17:05:39 - INFO - qdiff.optimization.model_recon -   Block 30 iter 950/1000: GPU 28.10GB
+01/28/2026 17:07:30 - INFO - qdiff.optimization.model_recon -   Block 30 optimization complete
+01/28/2026 17:07:30 - INFO - qdiff.optimization.model_recon - Processing WAN block 31/39
+01/28/2026 17:07:30 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 17:07:30 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 31...
+01/28/2026 17:09:49 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 17:09:49 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 17:09:49 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.07GB
+01/28/2026 17:09:51 - INFO - qdiff.optimization.model_recon -   Block 31 iter 0/1000: GPU 28.11GB
+01/28/2026 17:11:49 - INFO - qdiff.optimization.model_recon -   Block 31 iter 50/1000: GPU 28.11GB
+01/28/2026 17:13:43 - INFO - qdiff.optimization.model_recon -   Block 31 iter 100/1000: GPU 28.11GB
+01/28/2026 17:15:35 - INFO - qdiff.optimization.model_recon -   Block 31 iter 150/1000: GPU 28.11GB
+01/28/2026 17:17:27 - INFO - qdiff.optimization.model_recon -   Block 31 iter 200/1000: GPU 28.11GB
+01/28/2026 17:19:22 - INFO - qdiff.optimization.model_recon -   Block 31 iter 250/1000: GPU 28.11GB
+01/28/2026 17:21:15 - INFO - qdiff.optimization.model_recon -   Block 31 iter 300/1000: GPU 28.11GB
+01/28/2026 17:23:08 - INFO - qdiff.optimization.model_recon -   Block 31 iter 350/1000: GPU 28.11GB
+01/28/2026 17:25:01 - INFO - qdiff.optimization.model_recon -   Block 31 iter 400/1000: GPU 28.11GB
+01/28/2026 17:26:53 - INFO - qdiff.optimization.model_recon -   Block 31 iter 450/1000: GPU 28.11GB
+01/28/2026 17:28:46 - INFO - qdiff.optimization.model_recon -   Block 31 iter 500/1000: GPU 28.11GB
+01/28/2026 17:30:38 - INFO - qdiff.optimization.model_recon -   Block 31 iter 550/1000: GPU 28.11GB
+01/28/2026 17:32:30 - INFO - qdiff.optimization.model_recon -   Block 31 iter 600/1000: GPU 28.11GB
+01/28/2026 17:34:23 - INFO - qdiff.optimization.model_recon -   Block 31 iter 650/1000: GPU 28.11GB
+01/28/2026 17:36:15 - INFO - qdiff.optimization.model_recon -   Block 31 iter 700/1000: GPU 28.11GB
+01/28/2026 17:38:06 - INFO - qdiff.optimization.model_recon -   Block 31 iter 750/1000: GPU 28.11GB
+01/28/2026 17:39:58 - INFO - qdiff.optimization.model_recon -   Block 31 iter 800/1000: GPU 28.11GB
+01/28/2026 17:41:49 - INFO - qdiff.optimization.model_recon -   Block 31 iter 850/1000: GPU 28.11GB
+01/28/2026 17:43:40 - INFO - qdiff.optimization.model_recon -   Block 31 iter 900/1000: GPU 28.11GB
+01/28/2026 17:45:33 - INFO - qdiff.optimization.model_recon -   Block 31 iter 950/1000: GPU 28.11GB
+01/28/2026 17:47:24 - INFO - qdiff.optimization.model_recon -   Block 31 optimization complete
+01/28/2026 17:47:24 - INFO - qdiff.optimization.model_recon - Processing WAN block 32/39
+01/28/2026 17:47:24 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 17:47:24 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 32...
+01/28/2026 17:49:46 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 17:49:46 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 17:49:46 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.08GB
+01/28/2026 17:49:49 - INFO - qdiff.optimization.model_recon -   Block 32 iter 0/1000: GPU 28.13GB
+01/28/2026 17:51:41 - INFO - qdiff.optimization.model_recon -   Block 32 iter 50/1000: GPU 28.13GB
+01/28/2026 17:53:33 - INFO - qdiff.optimization.model_recon -   Block 32 iter 100/1000: GPU 28.13GB
+01/28/2026 17:55:25 - INFO - qdiff.optimization.model_recon -   Block 32 iter 150/1000: GPU 28.13GB
+01/28/2026 17:57:17 - INFO - qdiff.optimization.model_recon -   Block 32 iter 200/1000: GPU 28.13GB
+01/28/2026 17:59:09 - INFO - qdiff.optimization.model_recon -   Block 32 iter 250/1000: GPU 28.13GB
+01/28/2026 18:01:01 - INFO - qdiff.optimization.model_recon -   Block 32 iter 300/1000: GPU 28.13GB
+01/28/2026 18:02:53 - INFO - qdiff.optimization.model_recon -   Block 32 iter 350/1000: GPU 28.13GB
+01/28/2026 18:04:44 - INFO - qdiff.optimization.model_recon -   Block 32 iter 400/1000: GPU 28.13GB
+01/28/2026 18:06:37 - INFO - qdiff.optimization.model_recon -   Block 32 iter 450/1000: GPU 28.13GB
+01/28/2026 18:08:28 - INFO - qdiff.optimization.model_recon -   Block 32 iter 500/1000: GPU 28.13GB
+01/28/2026 18:10:20 - INFO - qdiff.optimization.model_recon -   Block 32 iter 550/1000: GPU 28.13GB
+01/28/2026 18:12:12 - INFO - qdiff.optimization.model_recon -   Block 32 iter 600/1000: GPU 28.13GB
+01/28/2026 18:14:04 - INFO - qdiff.optimization.model_recon -   Block 32 iter 650/1000: GPU 28.13GB
+01/28/2026 18:15:56 - INFO - qdiff.optimization.model_recon -   Block 32 iter 700/1000: GPU 28.13GB
+01/28/2026 18:17:47 - INFO - qdiff.optimization.model_recon -   Block 32 iter 750/1000: GPU 28.13GB
+01/28/2026 18:19:39 - INFO - qdiff.optimization.model_recon -   Block 32 iter 800/1000: GPU 28.13GB
+01/28/2026 18:21:30 - INFO - qdiff.optimization.model_recon -   Block 32 iter 850/1000: GPU 28.13GB
+01/28/2026 18:23:22 - INFO - qdiff.optimization.model_recon -   Block 32 iter 900/1000: GPU 28.13GB
+01/28/2026 18:25:13 - INFO - qdiff.optimization.model_recon -   Block 32 iter 950/1000: GPU 28.13GB
+01/28/2026 18:27:04 - INFO - qdiff.optimization.model_recon -   Block 32 optimization complete
+01/28/2026 18:27:04 - INFO - qdiff.optimization.model_recon - Processing WAN block 33/39
+01/28/2026 18:27:04 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 18:27:04 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 33...
+01/28/2026 18:29:30 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 18:29:30 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 18:29:30 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.10GB
+01/28/2026 18:29:32 - INFO - qdiff.optimization.model_recon -   Block 33 iter 0/1000: GPU 28.14GB
+01/28/2026 18:31:24 - INFO - qdiff.optimization.model_recon -   Block 33 iter 50/1000: GPU 28.14GB
+01/28/2026 18:33:17 - INFO - qdiff.optimization.model_recon -   Block 33 iter 100/1000: GPU 28.14GB
+01/28/2026 18:35:09 - INFO - qdiff.optimization.model_recon -   Block 33 iter 150/1000: GPU 28.14GB
+01/28/2026 18:37:01 - INFO - qdiff.optimization.model_recon -   Block 33 iter 200/1000: GPU 28.14GB
+01/28/2026 18:38:53 - INFO - qdiff.optimization.model_recon -   Block 33 iter 250/1000: GPU 28.14GB
+01/28/2026 18:40:44 - INFO - qdiff.optimization.model_recon -   Block 33 iter 300/1000: GPU 28.14GB
+01/28/2026 18:42:36 - INFO - qdiff.optimization.model_recon -   Block 33 iter 350/1000: GPU 28.14GB
+01/28/2026 18:44:27 - INFO - qdiff.optimization.model_recon -   Block 33 iter 400/1000: GPU 28.14GB
+01/28/2026 18:46:18 - INFO - qdiff.optimization.model_recon -   Block 33 iter 450/1000: GPU 28.14GB
+01/28/2026 18:48:10 - INFO - qdiff.optimization.model_recon -   Block 33 iter 500/1000: GPU 28.14GB
+01/28/2026 18:50:02 - INFO - qdiff.optimization.model_recon -   Block 33 iter 550/1000: GPU 28.14GB
+01/28/2026 18:51:54 - INFO - qdiff.optimization.model_recon -   Block 33 iter 600/1000: GPU 28.14GB
+01/28/2026 18:53:45 - INFO - qdiff.optimization.model_recon -   Block 33 iter 650/1000: GPU 28.14GB
+01/28/2026 18:55:37 - INFO - qdiff.optimization.model_recon -   Block 33 iter 700/1000: GPU 28.14GB
+01/28/2026 18:57:29 - INFO - qdiff.optimization.model_recon -   Block 33 iter 750/1000: GPU 28.14GB
+01/28/2026 18:59:22 - INFO - qdiff.optimization.model_recon -   Block 33 iter 800/1000: GPU 28.14GB
+01/28/2026 19:01:14 - INFO - qdiff.optimization.model_recon -   Block 33 iter 850/1000: GPU 28.14GB
+01/28/2026 19:03:06 - INFO - qdiff.optimization.model_recon -   Block 33 iter 900/1000: GPU 28.14GB
+01/28/2026 19:04:56 - INFO - qdiff.optimization.model_recon -   Block 33 iter 950/1000: GPU 28.14GB
+01/28/2026 19:06:47 - INFO - qdiff.optimization.model_recon -   Block 33 optimization complete
+01/28/2026 19:06:47 - INFO - qdiff.optimization.model_recon - Processing WAN block 34/39
+01/28/2026 19:06:47 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 19:06:47 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 34...
+01/28/2026 19:09:17 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 19:09:17 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 19:09:17 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.11GB
+01/28/2026 19:09:19 - INFO - qdiff.optimization.model_recon -   Block 34 iter 0/1000: GPU 28.16GB
+01/28/2026 19:11:11 - INFO - qdiff.optimization.model_recon -   Block 34 iter 50/1000: GPU 28.16GB
+01/28/2026 19:13:02 - INFO - qdiff.optimization.model_recon -   Block 34 iter 100/1000: GPU 28.16GB
+01/28/2026 19:14:53 - INFO - qdiff.optimization.model_recon -   Block 34 iter 150/1000: GPU 28.16GB
+01/28/2026 19:16:45 - INFO - qdiff.optimization.model_recon -   Block 34 iter 200/1000: GPU 28.16GB
+01/28/2026 19:18:36 - INFO - qdiff.optimization.model_recon -   Block 34 iter 250/1000: GPU 28.16GB
+01/28/2026 19:20:27 - INFO - qdiff.optimization.model_recon -   Block 34 iter 300/1000: GPU 28.16GB
+01/28/2026 19:22:19 - INFO - qdiff.optimization.model_recon -   Block 34 iter 350/1000: GPU 28.16GB
+01/28/2026 19:24:10 - INFO - qdiff.optimization.model_recon -   Block 34 iter 400/1000: GPU 28.16GB
+01/28/2026 19:26:01 - INFO - qdiff.optimization.model_recon -   Block 34 iter 450/1000: GPU 28.16GB
+01/28/2026 19:27:53 - INFO - qdiff.optimization.model_recon -   Block 34 iter 500/1000: GPU 28.16GB
+01/28/2026 19:29:45 - INFO - qdiff.optimization.model_recon -   Block 34 iter 550/1000: GPU 28.16GB
+01/28/2026 19:31:36 - INFO - qdiff.optimization.model_recon -   Block 34 iter 600/1000: GPU 28.16GB
+01/28/2026 19:33:28 - INFO - qdiff.optimization.model_recon -   Block 34 iter 650/1000: GPU 28.16GB
+01/28/2026 19:35:19 - INFO - qdiff.optimization.model_recon -   Block 34 iter 700/1000: GPU 28.16GB
+01/28/2026 19:37:11 - INFO - qdiff.optimization.model_recon -   Block 34 iter 750/1000: GPU 28.16GB
+01/28/2026 19:39:03 - INFO - qdiff.optimization.model_recon -   Block 34 iter 800/1000: GPU 28.16GB
+01/28/2026 19:40:54 - INFO - qdiff.optimization.model_recon -   Block 34 iter 850/1000: GPU 28.16GB
+01/28/2026 19:42:46 - INFO - qdiff.optimization.model_recon -   Block 34 iter 900/1000: GPU 28.16GB
+01/28/2026 19:44:38 - INFO - qdiff.optimization.model_recon -   Block 34 iter 950/1000: GPU 28.16GB
+01/28/2026 19:46:28 - INFO - qdiff.optimization.model_recon -   Block 34 optimization complete
+01/28/2026 19:46:28 - INFO - qdiff.optimization.model_recon - Processing WAN block 35/39
+01/28/2026 19:46:28 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 19:46:28 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 35...
+01/28/2026 19:49:02 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 19:49:02 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 19:49:02 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.13GB
+01/28/2026 19:49:04 - INFO - qdiff.optimization.model_recon -   Block 35 iter 0/1000: GPU 28.17GB
+01/28/2026 19:50:56 - INFO - qdiff.optimization.model_recon -   Block 35 iter 50/1000: GPU 28.17GB
+01/28/2026 19:52:50 - INFO - qdiff.optimization.model_recon -   Block 35 iter 100/1000: GPU 28.17GB
+01/28/2026 19:54:42 - INFO - qdiff.optimization.model_recon -   Block 35 iter 150/1000: GPU 28.17GB
+01/28/2026 19:56:33 - INFO - qdiff.optimization.model_recon -   Block 35 iter 200/1000: GPU 28.17GB
+01/28/2026 19:58:26 - INFO - qdiff.optimization.model_recon -   Block 35 iter 250/1000: GPU 28.17GB
+01/28/2026 20:00:18 - INFO - qdiff.optimization.model_recon -   Block 35 iter 300/1000: GPU 28.17GB
+01/28/2026 20:02:11 - INFO - qdiff.optimization.model_recon -   Block 35 iter 350/1000: GPU 28.17GB
+01/28/2026 20:04:04 - INFO - qdiff.optimization.model_recon -   Block 35 iter 400/1000: GPU 28.17GB
+01/28/2026 20:05:59 - INFO - qdiff.optimization.model_recon -   Block 35 iter 450/1000: GPU 28.17GB
+01/28/2026 20:07:58 - INFO - qdiff.optimization.model_recon -   Block 35 iter 500/1000: GPU 28.17GB
+01/28/2026 20:09:56 - INFO - qdiff.optimization.model_recon -   Block 35 iter 550/1000: GPU 28.17GB
+01/28/2026 20:11:54 - INFO - qdiff.optimization.model_recon -   Block 35 iter 600/1000: GPU 28.17GB
+01/28/2026 20:13:49 - INFO - qdiff.optimization.model_recon -   Block 35 iter 650/1000: GPU 28.17GB
+01/28/2026 20:15:42 - INFO - qdiff.optimization.model_recon -   Block 35 iter 700/1000: GPU 28.17GB
+01/28/2026 20:17:34 - INFO - qdiff.optimization.model_recon -   Block 35 iter 750/1000: GPU 28.17GB
+01/28/2026 20:19:27 - INFO - qdiff.optimization.model_recon -   Block 35 iter 800/1000: GPU 28.17GB
+01/28/2026 20:21:20 - INFO - qdiff.optimization.model_recon -   Block 35 iter 850/1000: GPU 28.17GB
+01/28/2026 20:23:13 - INFO - qdiff.optimization.model_recon -   Block 35 iter 900/1000: GPU 28.17GB
+01/28/2026 20:25:05 - INFO - qdiff.optimization.model_recon -   Block 35 iter 950/1000: GPU 28.17GB
+01/28/2026 20:26:55 - INFO - qdiff.optimization.model_recon -   Block 35 optimization complete
+01/28/2026 20:26:55 - INFO - qdiff.optimization.model_recon - Processing WAN block 36/39
+01/28/2026 20:26:55 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 20:26:55 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 36...
+01/28/2026 20:29:33 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 20:29:33 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 20:29:33 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.14GB
+01/28/2026 20:29:35 - INFO - qdiff.optimization.model_recon -   Block 36 iter 0/1000: GPU 28.19GB
+01/28/2026 20:31:27 - INFO - qdiff.optimization.model_recon -   Block 36 iter 50/1000: GPU 28.19GB
+01/28/2026 20:33:19 - INFO - qdiff.optimization.model_recon -   Block 36 iter 100/1000: GPU 28.19GB
+01/28/2026 20:35:10 - INFO - qdiff.optimization.model_recon -   Block 36 iter 150/1000: GPU 28.19GB
+01/28/2026 20:37:03 - INFO - qdiff.optimization.model_recon -   Block 36 iter 200/1000: GPU 28.19GB
+01/28/2026 20:38:54 - INFO - qdiff.optimization.model_recon -   Block 36 iter 250/1000: GPU 28.19GB
+01/28/2026 20:40:46 - INFO - qdiff.optimization.model_recon -   Block 36 iter 300/1000: GPU 28.19GB
+01/28/2026 20:42:38 - INFO - qdiff.optimization.model_recon -   Block 36 iter 350/1000: GPU 28.19GB
+01/28/2026 20:44:29 - INFO - qdiff.optimization.model_recon -   Block 36 iter 400/1000: GPU 28.19GB
+01/28/2026 20:46:21 - INFO - qdiff.optimization.model_recon -   Block 36 iter 450/1000: GPU 28.19GB
+01/28/2026 20:48:12 - INFO - qdiff.optimization.model_recon -   Block 36 iter 500/1000: GPU 28.19GB
+01/28/2026 20:50:03 - INFO - qdiff.optimization.model_recon -   Block 36 iter 550/1000: GPU 28.19GB
+01/28/2026 20:51:55 - INFO - qdiff.optimization.model_recon -   Block 36 iter 600/1000: GPU 28.19GB
+01/28/2026 20:53:46 - INFO - qdiff.optimization.model_recon -   Block 36 iter 650/1000: GPU 28.19GB
+01/28/2026 20:55:37 - INFO - qdiff.optimization.model_recon -   Block 36 iter 700/1000: GPU 28.19GB
+01/28/2026 20:57:30 - INFO - qdiff.optimization.model_recon -   Block 36 iter 750/1000: GPU 28.19GB
+01/28/2026 20:59:21 - INFO - qdiff.optimization.model_recon -   Block 36 iter 800/1000: GPU 28.19GB
+01/28/2026 21:01:13 - INFO - qdiff.optimization.model_recon -   Block 36 iter 850/1000: GPU 28.19GB
+01/28/2026 21:03:05 - INFO - qdiff.optimization.model_recon -   Block 36 iter 900/1000: GPU 28.19GB
+01/28/2026 21:04:57 - INFO - qdiff.optimization.model_recon -   Block 36 iter 950/1000: GPU 28.19GB
+01/28/2026 21:06:47 - INFO - qdiff.optimization.model_recon -   Block 36 optimization complete
+01/28/2026 21:06:47 - INFO - qdiff.optimization.model_recon - Processing WAN block 37/39
+01/28/2026 21:06:47 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 21:06:47 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 37...
+01/28/2026 21:09:29 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 21:09:29 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 21:09:29 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.16GB
+01/28/2026 21:09:32 - INFO - qdiff.optimization.model_recon -   Block 37 iter 0/1000: GPU 28.20GB
+01/28/2026 21:11:25 - INFO - qdiff.optimization.model_recon -   Block 37 iter 50/1000: GPU 28.20GB
+01/28/2026 21:13:17 - INFO - qdiff.optimization.model_recon -   Block 37 iter 100/1000: GPU 28.20GB
+01/28/2026 21:15:08 - INFO - qdiff.optimization.model_recon -   Block 37 iter 150/1000: GPU 28.20GB
+01/28/2026 21:17:00 - INFO - qdiff.optimization.model_recon -   Block 37 iter 200/1000: GPU 28.20GB
+01/28/2026 21:18:53 - INFO - qdiff.optimization.model_recon -   Block 37 iter 250/1000: GPU 28.20GB
+01/28/2026 21:20:45 - INFO - qdiff.optimization.model_recon -   Block 37 iter 300/1000: GPU 28.20GB
+01/28/2026 21:22:38 - INFO - qdiff.optimization.model_recon -   Block 37 iter 350/1000: GPU 28.20GB
+01/28/2026 21:24:31 - INFO - qdiff.optimization.model_recon -   Block 37 iter 400/1000: GPU 28.20GB
+01/28/2026 21:26:23 - INFO - qdiff.optimization.model_recon -   Block 37 iter 450/1000: GPU 28.20GB
+01/28/2026 21:28:15 - INFO - qdiff.optimization.model_recon -   Block 37 iter 500/1000: GPU 28.20GB
+01/28/2026 21:30:08 - INFO - qdiff.optimization.model_recon -   Block 37 iter 550/1000: GPU 28.20GB
+01/28/2026 21:32:01 - INFO - qdiff.optimization.model_recon -   Block 37 iter 600/1000: GPU 28.20GB
+01/28/2026 21:33:52 - INFO - qdiff.optimization.model_recon -   Block 37 iter 650/1000: GPU 28.20GB
+01/28/2026 21:35:44 - INFO - qdiff.optimization.model_recon -   Block 37 iter 700/1000: GPU 28.20GB
+01/28/2026 21:37:36 - INFO - qdiff.optimization.model_recon -   Block 37 iter 750/1000: GPU 28.20GB
+01/28/2026 21:39:28 - INFO - qdiff.optimization.model_recon -   Block 37 iter 800/1000: GPU 28.20GB
+01/28/2026 21:41:20 - INFO - qdiff.optimization.model_recon -   Block 37 iter 850/1000: GPU 28.20GB
+01/28/2026 21:43:12 - INFO - qdiff.optimization.model_recon -   Block 37 iter 900/1000: GPU 28.20GB
+01/28/2026 21:45:06 - INFO - qdiff.optimization.model_recon -   Block 37 iter 950/1000: GPU 28.20GB
+01/28/2026 21:46:56 - INFO - qdiff.optimization.model_recon -   Block 37 optimization complete
+01/28/2026 21:46:57 - INFO - qdiff.optimization.model_recon - Processing WAN block 38/39
+01/28/2026 21:46:57 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 21:46:57 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 38...
+01/28/2026 21:49:42 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 21:49:42 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 21:49:43 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.17GB
+01/28/2026 21:49:45 - INFO - qdiff.optimization.model_recon -   Block 38 iter 0/1000: GPU 28.22GB
+01/28/2026 21:51:39 - INFO - qdiff.optimization.model_recon -   Block 38 iter 50/1000: GPU 28.22GB
+01/28/2026 21:53:32 - INFO - qdiff.optimization.model_recon -   Block 38 iter 100/1000: GPU 28.22GB
+01/28/2026 21:55:25 - INFO - qdiff.optimization.model_recon -   Block 38 iter 150/1000: GPU 28.22GB
+01/28/2026 21:57:18 - INFO - qdiff.optimization.model_recon -   Block 38 iter 200/1000: GPU 28.22GB
+01/28/2026 21:59:10 - INFO - qdiff.optimization.model_recon -   Block 38 iter 250/1000: GPU 28.22GB
+01/28/2026 22:01:03 - INFO - qdiff.optimization.model_recon -   Block 38 iter 300/1000: GPU 28.22GB
+01/28/2026 22:02:56 - INFO - qdiff.optimization.model_recon -   Block 38 iter 350/1000: GPU 28.22GB
+01/28/2026 22:04:49 - INFO - qdiff.optimization.model_recon -   Block 38 iter 400/1000: GPU 28.22GB
+01/28/2026 22:06:42 - INFO - qdiff.optimization.model_recon -   Block 38 iter 450/1000: GPU 28.22GB
+01/28/2026 22:08:36 - INFO - qdiff.optimization.model_recon -   Block 38 iter 500/1000: GPU 28.22GB
+01/28/2026 22:10:29 - INFO - qdiff.optimization.model_recon -   Block 38 iter 550/1000: GPU 28.22GB
+01/28/2026 22:12:22 - INFO - qdiff.optimization.model_recon -   Block 38 iter 600/1000: GPU 28.22GB
+01/28/2026 22:14:14 - INFO - qdiff.optimization.model_recon -   Block 38 iter 650/1000: GPU 28.22GB
+01/28/2026 22:16:07 - INFO - qdiff.optimization.model_recon -   Block 38 iter 700/1000: GPU 28.22GB
+01/28/2026 22:17:58 - INFO - qdiff.optimization.model_recon -   Block 38 iter 750/1000: GPU 28.22GB
+01/28/2026 22:19:49 - INFO - qdiff.optimization.model_recon -   Block 38 iter 800/1000: GPU 28.22GB
+01/28/2026 22:21:41 - INFO - qdiff.optimization.model_recon -   Block 38 iter 850/1000: GPU 28.22GB
+01/28/2026 22:23:33 - INFO - qdiff.optimization.model_recon -   Block 38 iter 900/1000: GPU 28.22GB
+01/28/2026 22:25:25 - INFO - qdiff.optimization.model_recon -   Block 38 iter 950/1000: GPU 28.22GB
+01/28/2026 22:27:14 - INFO - qdiff.optimization.model_recon -   Block 38 optimization complete
+01/28/2026 22:27:14 - INFO - qdiff.optimization.model_recon - Processing WAN block 39/39
+01/28/2026 22:27:14 - INFO - qdiff.optimization.model_recon -   Limiting samples from 60 to 8 for per-block optimization
+01/28/2026 22:27:14 - INFO - qdiff.optimization.model_recon -   Collecting inputs/outputs for block 39...
+01/28/2026 22:30:04 - INFO - qdiff.optimization.model_recon -   Block input shape: torch.Size([8, 75600, 5120])
+01/28/2026 22:30:04 - INFO - qdiff.optimization.model_recon -   Block output shape: torch.Size([8, 75600, 5120])
+01/28/2026 22:30:04 - INFO - qdiff.optimization.model_recon -   GPU memory before block optimization: 28.19GB
+01/28/2026 22:30:06 - INFO - qdiff.optimization.model_recon -   Block 39 iter 0/1000: GPU 28.23GB
+01/28/2026 22:31:58 - INFO - qdiff.optimization.model_recon -   Block 39 iter 50/1000: GPU 28.23GB
+01/28/2026 22:33:50 - INFO - qdiff.optimization.model_recon -   Block 39 iter 100/1000: GPU 28.23GB
+01/28/2026 22:35:42 - INFO - qdiff.optimization.model_recon -   Block 39 iter 150/1000: GPU 28.23GB
+01/28/2026 22:37:34 - INFO - qdiff.optimization.model_recon -   Block 39 iter 200/1000: GPU 28.23GB
+01/28/2026 22:39:26 - INFO - qdiff.optimization.model_recon -   Block 39 iter 250/1000: GPU 28.23GB
+01/28/2026 22:41:18 - INFO - qdiff.optimization.model_recon -   Block 39 iter 300/1000: GPU 28.23GB
+01/28/2026 22:43:10 - INFO - qdiff.optimization.model_recon -   Block 39 iter 350/1000: GPU 28.23GB
+01/28/2026 22:45:02 - INFO - qdiff.optimization.model_recon -   Block 39 iter 400/1000: GPU 28.23GB
+01/28/2026 22:46:54 - INFO - qdiff.optimization.model_recon -   Block 39 iter 450/1000: GPU 28.23GB
+01/28/2026 22:48:46 - INFO - qdiff.optimization.model_recon -   Block 39 iter 500/1000: GPU 28.23GB
+01/28/2026 22:50:39 - INFO - qdiff.optimization.model_recon -   Block 39 iter 550/1000: GPU 28.23GB
+01/28/2026 22:52:32 - INFO - qdiff.optimization.model_recon -   Block 39 iter 600/1000: GPU 28.23GB
+01/28/2026 22:54:24 - INFO - qdiff.optimization.model_recon -   Block 39 iter 650/1000: GPU 28.23GB
+01/28/2026 22:56:17 - INFO - qdiff.optimization.model_recon -   Block 39 iter 700/1000: GPU 28.23GB
+01/28/2026 22:58:10 - INFO - qdiff.optimization.model_recon -   Block 39 iter 750/1000: GPU 28.23GB
+01/28/2026 23:00:03 - INFO - qdiff.optimization.model_recon -   Block 39 iter 800/1000: GPU 28.23GB
+01/28/2026 23:01:56 - INFO - qdiff.optimization.model_recon -   Block 39 iter 850/1000: GPU 28.23GB
+01/28/2026 23:03:47 - INFO - qdiff.optimization.model_recon -   Block 39 iter 900/1000: GPU 28.23GB
+01/28/2026 23:05:40 - INFO - qdiff.optimization.model_recon -   Block 39 iter 950/1000: GPU 28.23GB
+01/28/2026 23:07:31 - INFO - qdiff.optimization.model_recon -   Block 39 optimization complete
+01/28/2026 23:07:31 - INFO - qdiff.optimization.model_recon - Completed 40/40 blocks
+01/28/2026 23:07:31 - INFO - qdiff.optimization.model_recon - All 40 WAN blocks optimized
+01/28/2026 23:07:31 - INFO - __main__ - T2: Weight optimization complete!
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model - CHECKPOINT SAVE SUMMARY
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model -   Total keys saved: 2916
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model -   LoRA layers: 406
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model -   TQE frame weights: 240
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model -   Weight quantizer delta (scales): 406
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model -   Act quantizer scale: 0
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model -   Act quantizer act_scale (smooth quant): 406
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model -   BaseQuantizer buffers: 0
+01/28/2026 23:07:31 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:32 - INFO - __main__ - T2: No activation optimization params, skipping
+01/28/2026 23:07:33 - INFO - __main__ - T2: Validating quantizers before save...
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - QUANTIZER VALIDATION RESULTS
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Total QuantLayers: 406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   NVFP4 weight quantizers: 406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   NVFP4 act quantizers: 406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Weight quantizers with delta: 406/406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Weight quantizers init_done: 406/406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Act quantizers with scale: 0/406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Act quantizers with act_scale (smooth quant): 406/406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Act quantizers init_done: 406/406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:33 - INFO - __main__ - T2: Saving checkpoint...
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - CHECKPOINT SAVE SUMMARY
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Total keys saved: 2916
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   LoRA layers: 406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   TQE frame weights: 240
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Weight quantizer delta (scales): 406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Act quantizer scale: 0
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   Act quantizer act_scale (smooth quant): 406
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model -   BaseQuantizer buffers: 0
+01/28/2026 23:07:33 - INFO - qdiff.models.quant_model - ============================================================
+01/28/2026 23:07:34 - INFO - __main__ - T2: Saved checkpoint to quant_output/ckpt_t2.pth
+01/28/2026 23:07:34 - INFO - __main__ - T2: Verifying saved checkpoint...
+01/28/2026 23:07:34 - INFO - __main__ - T2: Checkpoint verification:
+01/28/2026 23:07:34 - INFO - __main__ -   Total keys: 2916
+01/28/2026 23:07:34 - INFO - __main__ -   LoRA keys: 1624
+01/28/2026 23:07:34 - INFO - __main__ -   TQE keys: 240
+01/28/2026 23:07:34 - INFO - __main__ -   Weight delta keys: 406
+01/28/2026 23:07:34 - INFO - __main__ -   Act scale keys: 0
+01/28/2026 23:07:34 - INFO - __main__ -   Smooth quant act_scale keys: 406
+01/28/2026 23:07:34 - INFO - __main__ - 
+Calibration complete! Checkpoints saved to quant_output/
+01/28/2026 23:07:34 - INFO - __main__ -   - ckpt_t2.pth: Transformer 2 quantization parameters