fix: CLI public release fallback to species_labels.json + safetensors

Public HF repo missing private data/wa_plants_200k/manifest CSVs and pt ckpts. Patch load_species_map to use species_labels.json, add _resolve_ckpt() for PlantDetect-*.safetensors, safetensors.torch.load_file support, Dense/MoE auto-detect, robust cmd_info. Tested: info + single/multi-view on RTX4050 BF16/FP8 AdaRound 104MB.

plant_cli.py

@@ -27,11 +27,31 @@ import torchvision.transforms as T
 from src.data.plant import IMAGENET_MEAN, IMAGENET_STD
 
 MANIFEST = BASE/"data/wa_plants_200k/manifest_for_train.csv"
+SPECIES_JSON = BASE/"species_labels.json"
 CKPT_VISUAL = BASE/"data/plant_phase3b_otherblue.pt"
+# Public HF release fallbacks (root *.safetensors) - used when private data/ ckpts missing
+PUBLIC_CKPTS = [
+    BASE/"PlantDetect-FP8-AdaRound.safetensors",
+    BASE/"PlantDetect-BF16.safetensors",
+    BASE/"PlantDetect-Dense-FP8-AdaRound.safetensors",
+    BASE/"PlantDetect-Dense-BF16.safetensors",
+    BASE/"PlantDetect-4View-FP8-AdaRound.safetensors",
+    BASE/"PlantDetect-4View-BF16.safetensors",
+    BASE/"PlantDetect-Dense-4View-FP8-AdaRound.safetensors",
+    BASE/"PlantDetect-Dense-4View-BF16.safetensors",
+]
 CKPT_FUSION = BASE/"data/plant_phase5_fusion.pt"
 E5_DIR = BASE/"data/plant_phase5_e5"
 CKPT_OLD = BASE/"data/plant_phase2_200k.pt"
 
+def _resolve_ckpt(preferred=CKPT_VISUAL):
+    if preferred and preferred.exists():
+        return preferred
+    for cand in PUBLIC_CKPTS:
+        if cand.exists():
+            return cand
+    return preferred
+
 # fix PIL large image
 Image.MAX_IMAGE_PIXELS = 300_000_000
 
@@ -41,6 +61,19 @@ def get_transform(img_size=336):
     return T.Compose([T.Resize(int(img_size*1.14)), T.CenterCrop(img_size), T.ToTensor(), T.Normalize(IMAGENET_MEAN, IMAGENET_STD)])
 
 def load_species_map(manifest=MANIFEST):
+    # Public release: species_labels.json (999 WA species) - preferred
+    if SPECIES_JSON.exists():
+        data=json.loads(SPECIES_JSON.read_text(encoding="utf-8"))
+        if "idx_to_species" in data:
+            i2s={int(k):v for k,v in data["idx_to_species"].items()}
+            spp=[i2s[i] for i in sorted(i2s)]
+            s2i={s:i for i,s in i2s.items()}
+            return s2i, i2s, spp
+        spp=data.get("species", [])
+        s2i={s:i for i,s in enumerate(spp)}
+        i2s={i:s for s,i in s2i.items()}
+        return s2i, i2s, spp
+    # Fallback: private manifest CSV (training)
     df=pd.read_csv(manifest)
     spp=sorted(df[df["status"]=="downloaded"]["species"].unique())
     s2i={s:i for i,s in enumerate(spp)}
@@ -70,10 +103,28 @@ def open_image_pil(path):
     return im.convert("RGB")
 
 def load_visual_model(ckpt=CKPT_VISUAL, n_classes=999, device="cuda"):
+    # auto-resolve to public safetensors if private ckpt missing
+    ckpt=_resolve_ckpt(ckpt)
     s2i,i2s,spp = load_species_map()
-    model=PlantViT(stem_name="vit_base_patch16_dinov3", n_classes=n_classes, use_moe=True, num_ffn=16).to(device)
+    # n_classes from species list (public 999) overrides arg if mismatch
+    if len(spp)!=n_classes:
+        n_classes=len(spp)
+    # Dense vs MoE auto-detect from ckpt name
+    is_dense = "Dense" in ckpt.name if ckpt else False
+    use_moe = None if is_dense else True
+    model=PlantViT(stem_name="vit_base_patch16_dinov3", n_classes=n_classes, use_moe=use_moe, num_ffn=16).to(device)
     if ckpt.exists():
-        sd=torch.load(ckpt, map_location=device)
+        if ckpt.suffix==".safetensors":
+            try:
+                from safetensors.torch import load_file
+                try:
+                    sd=load_file(str(ckpt), device=device)
+                except Exception:
+                    sd=load_file(str(ckpt))
+            except ImportError:
+                raise RuntimeError("safetensors required: pip install safetensors")
+        else:
+            sd=torch.load(ckpt, map_location=device)
         if isinstance(sd, dict) and "model" in sd: sd=sd["model"]
         model.load_state_dict(sd, strict=False)
         log(f"loaded {ckpt.name} {ckpt.stat().st_size/1e6:.1f}MB")
@@ -349,16 +400,41 @@ def cmd_batch(args):
 def cmd_info(args):
     print("WA Plant Identifier — Built with DINOv3")
     print("Models:")
-    for p in [CKPT_VISUAL, CKPT_FUSION, E5_DIR, MANIFEST]:
+    # show both private and public candidates
+    cand_paths = [_resolve_ckpt(CKPT_VISUAL)] + PUBLIC_CKPTS
+    seen=set()
+    for p in cand_paths:
+        if str(p) in seen: continue
+        seen.add(str(p))
         exists=p.exists()
+        try:
+            rel=p.relative_to(BASE)
+        except ValueError:
+            rel=p
         size=f"{p.stat().st_size/1e6:.1f}MB" if exists and p.is_file() else ("dir" if exists else "missing")
-        print(f"  {p.relative_to(BASE)}  {size}  {'OK' if exists else 'MISSING'}")
-    s2i,i2s,spp = load_species_map()
-    print(f"Species 999 ({len(spp)} loaded) e.g. {spp[0]}, {spp[10]}")
+        print(f"  {rel}  {size}  {'OK' if exists else 'MISSING'}")
+    for p in [CKPT_FUSION, E5_DIR, MANIFEST, SPECIES_JSON]:
+        exists=p.exists()
+        try:
+            rel=p.relative_to(BASE)
+        except ValueError:
+            rel=p
+        size=f"{p.stat().st_size/1e6:.1f}MB" if exists and p.is_file() else ("dir" if p.exists() and p.is_dir() else "missing")
+        print(f"  {rel}  {size}  {'OK' if exists else 'MISSING'}")
+    try:
+        s2i,i2s,spp = load_species_map()
+        print(f"Species {len(spp)} ({len(spp)} loaded) e.g. {spp[0]}, {spp[10] if len(spp)>10 else spp[-1]}")
+    except Exception as e:
+        print(f"Species map failed: {e}")
+        spp=[]
     print(f"Device {'cuda '+torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'cpu'} torch {torch.__version__}")
     print(f"Benchmark val 27673: visual 89.21% Top-1 96.19% Top-5 -> fusion with text 92.00% 97.51%")
-    print(f"Multi-view: 1-view 89.04% -> 4-view 99.33% (pseudo)")
+    print(f"Multi-view: 1-view 89.04% -> 4-view 99.33% (pseudo) | Public HF: MoE 89.31% / Dense 82.08% single, 99.42%/98.31% 4-view")
     print("\nLicensing: DINOv3 Meta commercial grant required (LICENSE.md), text e5 MIT, data GBIF per-image BY/BY-NC")
+    if _resolve_ckpt(CKPT_VISUAL).exists():
+        print(f"Active ckpt: {_resolve_ckpt(CKPT_VISUAL).name} ({_resolve_ckpt(CKPT_VISUAL).stat().st_size/1e6:.1f}MB)")
+    else:
+        print("Active ckpt: MISSING - add PlantDetect-*.safetensors to project root")
 
 def main():
     parser=argparse.ArgumentParser(description="WA Plant Identifier — Built with DINOv3 | DSLR/phone seamless CLI", formatter_class=argparse.RawTextHelpFormatter)