Complete FireViewer Android learning SDK, source and verification records

conversion/package/scripts/convert_onnx.py

@@ -0,0 +1,20 @@
+"""Fix batch/spatial dimensions explicitly before the direct FlatBuffer export."""
+import json,subprocess,sys
+from pathlib import Path
+import onnx
+
+source,out=map(Path,sys.argv[1:3])
+spatial_size=int(sys.argv[3])
+model=onnx.load(str(source),load_external_data=False)
+shapes=[]
+for inp in model.graph.input:
+ dims=[]
+ for axis,dim in enumerate(inp.type.tensor_type.shape.dim):
+  if dim.dim_value>0: dims.append(dim.dim_value)
+  elif axis==0:dims.append(1)
+  elif len(inp.type.tensor_type.shape.dim)==4 and axis in {2,3}:dims.append(spatial_size)
+  else:raise ValueError(f'Unresolved input dimension: {inp.name} axis={axis}')
+ shapes.append(inp.name+':'+','.join(map(str,dims)))
+out.mkdir(parents=True,exist_ok=True)
+(out/'input-shapes.json').write_text(json.dumps(shapes,indent=2)+'\n')
+subprocess.run([str(Path(sys.executable).parent/'onnx2tf'),'-i',str(source),'-o',str(out),'-b','1','-ois',*shapes],check=True)

reports/fireviewer-completion.json

@@ -1,5 +1,5 @@
 {
-  "generated_at": "2026-09-20T00:35:11.670200+00:00",
+  "generated_at": "2026-09-20T00:36:23.287838+00:00",
   "repo": "fireviewer/litert-models",
   "models": [
     {