Sample with the paper's Euler 100 steps; model card with the paper's final results.

README.md

@@ -39,13 +39,14 @@ model = resflow.load_pretrained()                     # this repository
 
 vols = model.generate('meander', n=8)                 # (8, 64, 64, 32) uint8, 1 = sand
 vols = model.generate('lobe', n=4, ntg=0.35, width_cells=30, azimuth=45)
+column = [0] * 10 + [1] * 8 + [0] * 14                # a well log, base to top: 1 sand, 0 mud, -1 unknown
 ens = model.generate('meander', n=32,                 # every realisation honours the well exactly
                      wells=[resflow.Well(x=32, y=32, facies=column)])
 field = model.generate_field('lobe', shape=(512, 512, 32), ntg=0.5)
 ```
 
 A walkthrough notebook is in the code repository at `examples/pretrained/quickstart.ipynb`. A GPU is recommended:
-on an NVIDIA GH200 a 64×64×32 volume takes about 1.6 s in batches of 128 and a 512×512×32 field about 4 minutes.
+on an NVIDIA GH200 a 64×64×32 volume takes about 0.8 s in batches of 128 and a 512×512×32 field about 2 minutes.
 A CPU works, but takes tens of minutes per volume.
 
 ## Inputs
@@ -73,7 +74,7 @@ Parameters left out take the environment's typical value (the median over its tr
 A 32.6M-parameter 3D diffusion transformer (4×4×2 patches, 12 blocks, width 384, 6 heads, QK-norm, 3D rotary
 positions with θ = 100, adaptive layer norm) trained with conditional flow matching. The 18-number condition vector
 holds the environment one-hot, NTG, width, depth, the sine and cosine of the azimuth, and five environment-specific
-parameters; wells enter as two extra input channels. Sampling uses Heun's method with 100 steps and
+parameters; wells enter as two extra input channels. Sampling uses Euler's method with 100 steps and
 classifier-free guidance 3. Volumes use global attention as in training; fields switch to sliding-window attention
 of ±12 tokens per axis, so any extent is generated whole, without tiling.
 
@@ -84,10 +85,12 @@ the final epoch-40 exponential moving average.
 ## Evaluation
 
 On ResBench, which scores each check against the simulator's own run-to-run variability (1 means
-indistinguishable from the simulator, lower is better), ResFlow scores 2.56 overall: 2.33 on unconditional
-volumes, 3.12 on well-conditioned volumes, and 2.24 on fields. Given the same inputs, it is closer to the simulator
-than GANSim-3D, DiffSIM and a 3D latent diffusion model in every environment where they were compared, and it beats single-environment specialists with its own
-architecture and recipe in six of eight environments. Details are in the paper.
+indistinguishable from the simulator, lower is better), ResFlow scores 1.70 overall: 1.24 on unconditional
+volumes, 1.96 on well-conditioned volumes, and 1.90 on fields. It beats single-environment specialists with its own
+architecture and recipe in seven of eight environments. Given the same inputs, it is far closer to the simulator
+than GANSim-3D and a 3D latent diffusion model, and on par with DiffSIM, a diffusion model five times larger that
+is trained per environment and not built for heterogeneous fields (meander 1.67 against 1.81, PV shoestring 1.55
+against 1.52, SH proximal 1.84 against 1.67). Details are in the paper.
 
 ## Limitations
 
@@ -97,8 +100,8 @@ architecture and recipe in six of eight environments. Details are in the paper.
   (the API warns).
 - Under a fixed input it underestimates the simulator's uncertainty in most channel environments, so its ensembles
   understate uncertainty there.
-- Distal sheets contain isolated sand specks; at field scale, connectivity in shoestring and delta fields deviates
-  most from the simulator.
+- At field scale, connectivity and variograms deviate most from the simulator, in shoestring, delta and SH proximal
+  fields, and lobe fields carry more isolated sand specks than the simulator's.
 
 ## Citation
 

config.json

@@ -351,7 +351,7 @@
   }
  },
  "sampler": {
-  "solver": "heun",
+  "solver": "euler",
   "steps": 100,
   "guidance": 3.0,
   "field_radius": 12