Replace dynamic-range int8 with weight-only int8

Same change as efficientnet_b1 (merged earlier today), applied for consistency across the family: this repo's dynamic-range int8 is only mildly affected, but weight-only int8 is strictly closer to the float model at the same size reduction.

Measured against the torch float reference on real photos plus fixed random tensors at native resolution: dynamic min logit correlation 0.971 (top-1 4/6) vs weight-only 1.000 (top-1 6/6). The replacement file is quantized directly from this repository's published float32 file (litert quantize, recipe weight_only_wi8_afp32), so the weights are identical.

The model card is updated in the same commit: the metric entries for the removed dynamic file are dropped and a short quantized-variant note with the measured numbers is added.

README.md

@@ -28,12 +28,6 @@ model-index:
           - name: Top 5 Accuracy (Full Precision)
             type: accuracy
             value:  0.9691
-          - name: Top 1 Accuracy (Dynamic Quantized wi8 afp32)
-            type: accuracy
-            value:  0.8383
-          - name: Top 5 Accuracy (Dynamic Quantized wi8 afp32)
-            type: accuracy
-            value:  0.9677
 ---
 
 # EfficientNet B6
@@ -49,6 +43,16 @@ acc@1 (on ImageNet-1K): 84.008%
 acc@5 (on ImageNet-1K): 96.916%    
 num_params: 43,040,704
 
+### Quantized variant
+
+`efficientnet_b6_weight_only_wi8_afp32.tflite` is a weight-only int8
+quantization of the same weights (about 3.7x smaller than float32).
+Weight-only quantization is used instead of dynamic-range quantization
+because EfficientNet's SE and SiLU layers are sensitive to activation
+quantization; in a spot check against the float model the weight-only
+file keeps the top-1 predictions on real photos with a minimum logit
+correlation of 1.000.
+
 ## Intended uses & limitations
 
 The model files were converted from pretrained weights from PyTorch Vision. The models may have their own licenses or terms and conditions derived from PyTorch Vision and the dataset used for training. It is your responsibility to determine whether you have permission to use the models for your use case.

efficientnet_b6_dynamic_wi8_afp32.tflite → efficientnet_b6_weight_only_wi8_afp32.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a652ddf1695371800b210de68c36a46aa7eebd8828d2ea1dbf6a790d2f193223
-size 46132584
+oid sha256:ca7a772019bcd4817d776902b4a7d41f4db154dcaa13ec5b75d9a92dc6a91dfc
+size 46252896