Publish MOSS-Image SFT AnyFlow recipes, adapters, and validation previews

.gitattributes

@@ -473,3 +473,19 @@ validation_images/step_17500_portrait_0_512x512.png filter=lfs diff=lfs merge=lf
 validation_images/step_17500_portrait_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
 validation_images/step_17500_text_0_512x512.png filter=lfs diff=lfs merge=lfs -text
 validation_images/step_17500_text_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_character_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_character_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_photo_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_photo_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_portrait_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_portrait_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_text_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-18000/assets/step_18000_text_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_character_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_character_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_photo_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_photo_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_portrait_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_portrait_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_text_0_512x512.png filter=lfs diff=lfs merge=lfs -text
+validation_images/step_18000_text_1_1024x1024.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -12,7 +12,7 @@ tags:
 
 Forward MeanFlow pretraining. Stage 2 initializes from the final EMA adapter.
 
-Status: **training**. Observed updates: 17,570 / 30,000.
+Status: **training**. Observed updates: 18,130 / 30,000.
 
 ## Training recipe
 
@@ -45,11 +45,11 @@ Schedule correction: the stage-1 run resumed from update 1,000 with the [flow-sc
 
 [Pre-correction previews](pre-schedule-fix/validation_images) remain available.
 
-[1,000](checkpoint-1000), [2,000](checkpoint-2000), [3,000](checkpoint-3000), [4,000](checkpoint-4000), [5,000](checkpoint-5000), [6,000](checkpoint-6000), [7,000](checkpoint-7000), [8,000](checkpoint-8000), [9,000](checkpoint-9000), [10,000](checkpoint-10000), [11,000](checkpoint-11000), [12,000](checkpoint-12000), [13,000](checkpoint-13000), [14,000](checkpoint-14000), [15,000](checkpoint-15000), [16,000](checkpoint-16000), [17,000](checkpoint-17000)
+[1,000](checkpoint-1000), [2,000](checkpoint-2000), [3,000](checkpoint-3000), [4,000](checkpoint-4000), [5,000](checkpoint-5000), [6,000](checkpoint-6000), [7,000](checkpoint-7000), [8,000](checkpoint-8000), [9,000](checkpoint-9000), [10,000](checkpoint-10000), [11,000](checkpoint-11000), [12,000](checkpoint-12000), [13,000](checkpoint-13000), [14,000](checkpoint-14000), [15,000](checkpoint-15000), [16,000](checkpoint-16000), [17,000](checkpoint-17000), [18,000](checkpoint-18000)
 
 [All validation previews](validation_images) include live and EMA comparisons at both resolutions.
 
-Latest live adapter: [update 17,000](checkpoint-17000/pytorch_lora_weights.safetensors).
+Latest live adapter: [update 18,000](checkpoint-18000/pytorch_lora_weights.safetensors).
 
 ## Inference
 

checkpoint-18000/ema/pytorch_lora_weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c626b35d60e7342d3615b693052d4701aaf0003dfd2be04489e71adeb568786
+size 294180952

checkpoint-18000/pytorch_lora_weights.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a10d56200d26d9a843041ec8c9a02f2bccbf279245520fe5b2221ac8fd9441bc
+size 294180952

publication.json

@@ -67,10 +67,14 @@
     "17000": {
       "live": "bf8bebb562f270654b867b209e1952da1c61e96a931cc03c9506da1973abc7a4",
       "ema": "c0088344e800a35db9517e0cf7fb8eb03b878a9b4a98a8ef5c1a7d1d65c4145f"
+    },
+    "18000": {
+      "live": "a10d56200d26d9a843041ec8c9a02f2bccbf279245520fe5b2221ac8fd9441bc",
+      "ema": "5c626b35d60e7342d3615b693052d4701aaf0003dfd2be04489e71adeb568786"
     }
   },
   "state": "training",
-  "step": 17570,
+  "step": 18130,
   "previews": [
     "validation_images/step_0_character_0_512x512.png",
     "validation_images/step_0_character_1_1024x1024.png",
@@ -224,6 +228,14 @@
     "validation_images/step_17500_portrait_1_1024x1024.png",
     "validation_images/step_17500_text_0_512x512.png",
     "validation_images/step_17500_text_1_1024x1024.png",
+    "validation_images/step_18000_character_0_512x512.png",
+    "validation_images/step_18000_character_1_1024x1024.png",
+    "validation_images/step_18000_photo_0_512x512.png",
+    "validation_images/step_18000_photo_1_1024x1024.png",
+    "validation_images/step_18000_portrait_0_512x512.png",
+    "validation_images/step_18000_portrait_1_1024x1024.png",
+    "validation_images/step_18000_text_0_512x512.png",
+    "validation_images/step_18000_text_1_1024x1024.png",
     "validation_images/step_2000_character_0_512x512.png",
     "validation_images/step_2000_character_1_1024x1024.png",
     "validation_images/step_2000_photo_0_512x512.png",