stupidog04 commited on
Commit
45c38e5
1 Parent(s): 84fed5c

commit files to HF hub

Browse files
README.md CHANGED
@@ -2,7 +2,6 @@
2
  tags:
3
  - image-classification
4
  - pytorch
5
- - huggingpics
6
  library_name: generic
7
  metrics:
8
  - accuracy
 
2
  tags:
3
  - image-classification
4
  - pytorch
 
5
  library_name: generic
6
  metrics:
7
  - accuracy
config.json CHANGED
@@ -6,13 +6,14 @@
6
  "attention_probs_dropout_prob": 0.0,
7
  "custom_pipelines": {
8
  "pair-classification": {
9
- "impl": "pair_classification.PairClassificationPipeline",
10
  "pt": [
11
  "ViTForImageClassification"
12
  ],
13
  "tf": [
14
  "TFViTForImageClassification"
15
- ]
 
16
  }
17
  },
18
  "encoder_stride": 16,
@@ -31,12 +32,12 @@
31
  "initializer_range": 0.02,
32
  "intermediate_size": 3072,
33
  "label2id": {
34
- "chk1_fail": 0,
35
- "chk1_pass": 1,
36
- "chk2_fail": 2,
37
- "chk2_pass": 3,
38
- "chk3_fail": 4,
39
- "chk3_pass": 5
40
  },
41
  "layer_norm_eps": 1e-12,
42
  "model_type": "vit",
 
6
  "attention_probs_dropout_prob": 0.0,
7
  "custom_pipelines": {
8
  "pair-classification": {
9
+ "impl": "pair_classification_pipeline.PairClassificationPipeline",
10
  "pt": [
11
  "ViTForImageClassification"
12
  ],
13
  "tf": [
14
  "TFViTForImageClassification"
15
+ ],
16
+ "type": "image"
17
  }
18
  },
19
  "encoder_stride": 16,
 
32
  "initializer_range": 0.02,
33
  "intermediate_size": 3072,
34
  "label2id": {
35
+ "chk1_fail": "0",
36
+ "chk1_pass": "1",
37
+ "chk2_fail": "2",
38
+ "chk2_pass": "3",
39
+ "chk3_fail": "4",
40
+ "chk3_pass": "5"
41
  },
42
  "layer_norm_eps": 1e-12,
43
  "model_type": "vit",
pair_classification.py → pair_classification_pipeline.py RENAMED
@@ -1,8 +1,35 @@
1
  from torchvision import transforms
 
2
  from transformers import ImageClassificationPipeline
3
  import torch
4
 
5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  class PairClassificationPipeline(ImageClassificationPipeline):
7
  pipe_to_tensor = transforms.ToTensor()
8
  pipe_to_pil = transforms.ToPILImage()
 
1
  from torchvision import transforms
2
+ from transformers import ViTFeatureExtractor, ViTForImageClassification
3
  from transformers import ImageClassificationPipeline
4
  import torch
5
 
6
 
7
+ class PreTrainedPipeline():
8
+ def __init__(self, path):
9
+ """
10
+ Initialize model
11
+ """
12
+ # self.processor = feature_extractor = ViTFeatureExtractor.from_pretrained(model_flag)
13
+ model_flag = 'google/vit-base-patch16-224-in21k'
14
+ # model_flag = 'google/vit-base-patch16-384'
15
+ self.feature_extractor = ViTFeatureExtractor.from_pretrained(model_flag)
16
+ self.model = ViTForImageClassification.from_pretrained(path)
17
+ self.pipe = PairClassificationPipeline(self.model, feature_extractor=self.feature_extractor)
18
+
19
+ def __call__(self, inputs):
20
+ """
21
+ Args:
22
+ inputs (:obj:`np.array`):
23
+ The raw waveform of audio received. By default at 16KHz.
24
+ Return:
25
+ A :obj:`dict`:. The object return should be liked {"text": "XXX"} containing
26
+ the detected text from the input audio.
27
+ """
28
+ # input_values = self.processor(inputs, return_tensors="pt", sampling_rate=self.sampling_rate).input_values # Batch size 1
29
+ # logits = self.model(input_values).logits.cpu().detach().numpy()[0]
30
+ return self.pipe(inputs)
31
+
32
+
33
  class PairClassificationPipeline(ImageClassificationPipeline):
34
  pipe_to_tensor = transforms.ToTensor()
35
  pipe_to_pil = transforms.ToPILImage()
pipeline.py CHANGED
@@ -1,36 +1,8 @@
1
  from torchvision import transforms
2
- from transformers import ViTFeatureExtractor, ViTForImageClassification, ViTConfig
3
  from transformers import ImageClassificationPipeline
4
  import torch
5
 
6
 
7
-
8
- class PreTrainedPipeline():
9
- def __init__(self, path):
10
- """
11
- Initialize model
12
- """
13
- # self.processor = feature_extractor = ViTFeatureExtractor.from_pretrained(model_flag)
14
- model_flag = 'google/vit-base-patch16-224-in21k'
15
- # model_flag = 'google/vit-base-patch16-384'
16
- self.feature_extractor = ViTFeatureExtractor.from_pretrained(model_flag)
17
- self.model = ViTForImageClassification.from_pretrained(path)
18
- self.pipe = PairClassificationPipeline(self.model, feature_extractor=self.feature_extractor)
19
-
20
- def __call__(self, inputs):
21
- """
22
- Args:
23
- inputs (:obj:`np.array`):
24
- The raw waveform of audio received. By default at 16KHz.
25
- Return:
26
- A :obj:`dict`:. The object return should be liked {"text": "XXX"} containing
27
- the detected text from the input audio.
28
- """
29
- # input_values = self.processor(inputs, return_tensors="pt", sampling_rate=self.sampling_rate).input_values # Batch size 1
30
- # logits = self.model(input_values).logits.cpu().detach().numpy()[0]
31
- return self.pipe(inputs)
32
-
33
-
34
  class PairClassificationPipeline(ImageClassificationPipeline):
35
  pipe_to_tensor = transforms.ToTensor()
36
  pipe_to_pil = transforms.ToPILImage()
 
1
  from torchvision import transforms
 
2
  from transformers import ImageClassificationPipeline
3
  import torch
4
 
5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  class PairClassificationPipeline(ImageClassificationPipeline):
7
  pipe_to_tensor = transforms.ToTensor()
8
  pipe_to_pil = transforms.ToPILImage()
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b5233e44bcfb172d9200f5f0c63a00ac4411a1783bf6dd42b48dce384a468da1
3
  size 345635761
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a709dd954e0bcc97bd8ed5652becb0709ac5ebc4d36f6e1ae7b15de08019dc01
3
  size 345635761
runs/events.out.tfevents.1666972137.sa103.11178.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93ee6a8b55c58e1a4f81c0e7b2484cc7986e796c8cad905ef2417c30f5469136
3
+ size 551