update

.zeno_cache/EMBEDDINGIMAGENET1K_V1.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8e0f69f152f195cc60e2967c00cde6b17a718412285ac5f215b544e27bcefee9
-size 32459378
+oid sha256:96927109662aeedd9bc8f360bac3f554f4a09f3cd36de6d4c0fcc40a6ad9a351
+size 6408501

.zeno_cache/EMBEDDINGIMAGENET1K_V2.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:96863b62f3d9ff89f195d8e9342943df694682f0026c9120b4f27d47f2bb6810
-size 32459378
+oid sha256:81ebfff49566d4a53d83837a1e77033fecd5842db1f2c5f387c49c75401ab715
+size 32459336

.zeno_cache/EMBEDDINGresnet+clip.pickle

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f9d66f744de0b18b789b09e80a3114a532e237d70ae5a7783fd16e78c60172b0
-size 1762825

.zeno_cache/OUTPUTIMAGENET1K_V1.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:af2452401beb86800c4331c61ae01702b9a797b638acb46ce214dbda6d1baa7d
-size 185742
+oid sha256:85b1ac79bc95a1ad26887539d789f9d023f9494fd0f627d91104c6fda4678268
+size 186098

.zeno_cache/OUTPUTIMAGENET1K_V2.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:929dc46aa2c0dea6466fccf1fbda55b42c11a48c20782b90089015eb698e5e2a
-size 185100
+oid sha256:61ab659b672b662ef69f6f785e6012da413e08df68196c8d0747a0736851a884
+size 185058

.zeno_cache/POSTDISTILLcorrectIMAGENET1K_V1.pickle

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:56cc4b148a91afe661b1b64b71546d4f929b108b4dd3675cebd6ecde9b540b33
-size 167854

.zeno_cache/POSTDISTILLcorrectIMAGENET1K_V2.pickle

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b58173c01dada0075defbbfa081b673693168fd560c2d9bbb8034022bea06f9f
-size 167854

.zeno_cache/{PREDISTILLblue_border_count.pickle → POSTDISTILLincorrectIMAGENET1K_V1.pickle}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cfcb503e51335ab308c35241a76bbfbc5fbd79b9e34c6c4024cbaf872449a7d6
-size 174191
+oid sha256:5fa0582cfdaca3982c08905933bdeacdbe56f0015b2c8bb5246ac78edf4a31f9
+size 167814

.zeno_cache/{OUTPUTresnet+clip.pickle → POSTDISTILLincorrectIMAGENET1K_V2.pickle}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:87c9f035a1876086a5f0aa4df4498a462f78d18de5323dd64a40fae989254057
-size 183793
+oid sha256:ee16c9cd7b0bc39b53fd7fabb4a2222a031c574d0266fc73179b24cdea7e4bc7
+size 167814

.zeno_cache/POSTDISTILLoutput_labelIMAGENET1K_V1.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6160edcb161d72d057042d5a7432d70669616282cfbc1771bb3d9da662389556
-size 185759
+oid sha256:3e7c313894d5f63d5e2f7778cd0295fd09ce7bc902965e0674ce2a7744323d2d
+size 186081

.zeno_cache/POSTDISTILLoutput_labelIMAGENET1K_V2.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:be315d03cd6119216109d06a93162857579ee0e96e7e3e41af4629d314afaecf
-size 185117
+oid sha256:cbdb8582997ccb184cb0b7e412874154ded3f0368d9b7224d449b194faa0f302
+size 185075

.zeno_cache/PREDISTILLborder_brightness.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3f6fb0629a3d26cf27b5c0f6b07f46cece9018db312c18e50893f21a50a4e73d
-size 199259
+oid sha256:ea251ea165919ebdbc1a8e6c7957d4b93fabc9a99167daa095b538ecebf46489
+size 199217

.zeno_cache/PREDISTILLbrightness.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0ad728441adc8c7e7d214f029ac7c4e505c40a34800292115736c61572653ad1
-size 199252
+oid sha256:1e1e21f96594b4d4dca3d641cc8b691326528f57918a161fadf9b0898efffd02
+size 199210

.zeno_cache/PREDISTILLred_count.pickle

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:e6a63cb93322c957969f428adf797b5fa2a73a03ecd4409102e8c1430d7e5cdd
-size 172023

.zeno_cache/folders.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ec0a6ccf9debf1c16781445c4b9106080d00478b0559469336db7c7b7b9711c8
-size 5
+oid sha256:e42dc27b16c2eea3b301e17de2ec138708fda5ecb1816365e3526c3d5ca58cd2
+size 42

.zeno_cache/reports.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:51652ef327bf266a3808c96bb7ed28087e097300962a18776b00e82abfe57c56
-size 353
+oid sha256:139a92c5150957a37e5ef2a4a5aa44ed8c6da6a133e98a3c738ae66e0084b844
+size 24190

.zeno_cache/slices.pickle

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:80f7becfc692f1cb3805bb29f5f8a24f9fead5399f045c7cc265dea7cd48eaa2
-size 846
+oid sha256:502150288ebdafdbc027ad2ac1e336e015d9368198debd7fc6d8674157d5fb75
+size 8496

color.py

@@ -1,51 +0,0 @@
-import colorsys
-import os
-
-import numpy as np
-from PIL import Image
-from zeno import distill
-from zeno.api import ZenoOptions
-
-
-def red_pixels(im):
-    arr = np.array(im)
-    count_red = 0
-    for x in range(arr.shape[0]):
-        for y in range(arr.shape[1]):
-            if arr[x, y, 0] > 180 and arr[x, y, 1] < 70 and arr[x, y, 2] < 70:
-                count_red += 1
-    return count_red
-
-
-def blue_border_pixels(im):
-    arr = np.array(im)
-    count_blue = 0
-    for x in range(arr.shape[0]):
-        for y in range(10):
-            hsv = colorsys.rgb_to_hsv(arr[x, y, 0], arr[x, y, 1], arr[x, y, 2])
-            if hsv[0] > 0.51 and hsv[0] < 0.72:
-                count_blue += 1
-    for x in range(arr.shape[0]):
-        for y in range(arr.shape[1] - 10, arr.shape[1]):
-            hsv = colorsys.rgb_to_hsv(arr[x, y, 0], arr[x, y, 1], arr[x, y, 2])
-            if hsv[0] > 0.51 and hsv[0] < 0.72:
-                count_blue += 1
-    return count_blue
-
-
-@distill
-def red_count(df, ops: ZenoOptions):
-    imgs = [
-        Image.open(os.path.join(ops.data_path, img)).convert("RGB")
-        for img in df[ops.data_column]
-    ]
-    return [red_pixels(im) for im in imgs]
-
-
-@distill
-def blue_border_count(df, ops):
-    imgs = [
-        Image.open(os.path.join(ops.data_path, img)).convert("RGB")
-        for img in df[ops.data_column]
-    ]
-    return [blue_border_pixels(im) for im in imgs]

config.toml

@@ -1,11 +1,14 @@
-functions = "./"
+functions = "./functions"
 view = "image-classification"
 metadata = "./imagenette_formatted.csv"
 models = ["IMAGENET1K_V1", "IMAGENET1K_V2"]
 data_path = "./imagenette/"
 data_column = "id"
+id_column = "id"
 label_column = "label"
 batch_size = 32
-port = 7860 
+port = 7860
 host = "0.0.0.0"
-editable = false
+editable = false
+# host = "localhost"
+# editable = true

brightness.py → functions/brightness.py

@@ -2,6 +2,7 @@ import os
 
 from PIL import Image
 from zeno import distill
+from zeno.api import DistillReturn
 
 
 def get_brightness(im):
@@ -35,10 +36,10 @@ def get_border_brightness(im):
 @distill
 def brightness(df, ops):
     imgs = [Image.open(os.path.join(ops.data_path, img)) for img in df[ops.data_column]]
-    return [get_brightness(im) for im in imgs]
+    return DistillReturn(distill_output=[get_brightness(im) for im in imgs])
 
 
 @distill
 def border_brightness(df, ops):
     imgs = [Image.open(os.path.join(ops.data_path, img)) for img in df[ops.data_column]]
-    return [get_border_brightness(im) for im in imgs]
+    return DistillReturn(distill_output=[get_border_brightness(im) for im in imgs])

loading.py → functions/loading.py

@@ -8,6 +8,7 @@ import torch
 
 
 from zeno import model
+from zeno.api import ModelReturn
 
 DEVICE = "cpu"
 
@@ -77,7 +78,7 @@ def load_model(model_path):
         batches_highest_confidence = outputs.argmax(dim=1)
         preds = [simple_labels[i] for i in batches_highest_confidence]
 
-        return preds, embeddings
+        return ModelReturn(model_output=preds, embedding=embeddings)
 
     return pred
 

functions/metrics.py

@@ -0,0 +1,40 @@
+from pandas import DataFrame
+from sklearn.metrics import f1_score, recall_score
+from zeno import ZenoOptions, MetricReturn, metric, distill, DistillReturn
+
+
+@metric
+def accuracy(df, ops: ZenoOptions):
+    if len(df) == 0:
+        return MetricReturn(metric=0)
+    return MetricReturn(
+        metric=100 * (df[ops.label_column] == df[ops.output_column]).sum() / len(df)
+    )
+
+
+@metric
+def recall(df, ops: ZenoOptions):
+    rec = recall_score(df[ops.label_column], df[ops.output_column], average="macro")
+    if type(rec) == float:
+        return MetricReturn(metric=100 * float(rec))
+    else:
+        return MetricReturn(metric=0)
+
+
+@metric
+def f1(df, ops: ZenoOptions):
+    f = f1_score(df[ops.label_column], df[ops.output_column], average="macro")
+    if type(f) == float:
+        return MetricReturn(metric=100 * f)
+    else:
+        return MetricReturn(metric=0)
+
+
+@distill
+def incorrect(df: DataFrame, ops: ZenoOptions):
+    return DistillReturn(distill_output=df[ops.label_column] != df[ops.output_column])
+
+
+@distill
+def output_label(df: DataFrame, ops: ZenoOptions):
+    return DistillReturn(distill_output=df[ops.output_column])

inference.py

@@ -1,11 +0,0 @@
-from zeno import inference, ZenoOptions
-import gradio as gr
-
-
-@inference
-def gradio_inference(ops: ZenoOptions):
-    return (
-        [gr.Image(type="filepath"), gr.Textbox(label="Label")],
-        gr.Text(label="Output"),
-        [ops.data_column, ops.label_column],
-    )

metrics.py

@@ -1,31 +0,0 @@
-from sklearn.metrics import f1_score, recall_score
-from zeno import ZenoOptions, distill, metric
-
-
-@metric
-def accuracy(df, ops: ZenoOptions):
-    if len(df) == 0:
-        return 0
-    return 100 * (df[ops.label_column] == df[ops.output_column]).sum() / len(df)
-
-
-@metric
-def recall(df, ops: ZenoOptions):
-    return 100 * recall_score(
-        df[ops.label_column], df[ops.output_column], average="macro"
-    )
-
-
-@metric
-def f1(df, ops: ZenoOptions):
-    return 100 * f1_score(df[ops.label_column], df[ops.output_column], average="macro")
-
-
-@distill
-def correct(df, ops: ZenoOptions):
-    return (df[ops.label_column] == df[ops.output_column]).tolist()
-
-
-@distill
-def output_label(df, ops: ZenoOptions):
-    return df[ops.output_column]

requirements.txt

@@ -1,4 +1,4 @@
-zenoml>=0.3.9
+zenoml>=0.4.6
 Pillow
 numpy
 torch