Datasets
Models
Downloads: 1
Diff of
/BioSeqNet/resnest/gluon/ablation.py
[000000]
..
[1fc74a]
Switch to unified view
| a | b/BioSeqNet/resnest/gluon/ablation.py | ||
|---|---|---|---|
| 1 | ##+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ |
||
| 2 | ## Created by: Hang Zhang |
||
| 3 | ## Email: zhanghang0704@gmail.com |
||
| 4 | ## Copyright (c) 2020 |
||
| 5 | ## |
||
| 6 | ## LICENSE file in the root directory of this source tree |
||
| 7 | ##+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ |
||
| 8 | |||
| 9 | """Ablation Study Models for ResNeSt""" |
||
| 10 | from .resnet import ResNet, Bottleneck |
||
| 11 | from mxnet import cpu |
||
| 12 | |||
| 13 | __all__ = ['resnest50_fast_1s1x64d', 'resnest50_fast_2s1x64d', 'resnest50_fast_4s1x64d', |
||
| 14 | 'resnest50_fast_1s2x40d', 'resnest50_fast_2s2x40d', 'resnest50_fast_4s2x40d', |
||
| 15 | 'resnest50_fast_1s4x24d'] |
||
| 16 | |||
| 17 | def resnest50_fast_1s1x64d(pretrained=False, root='~/.mxnet/models', ctx=cpu(0), **kwargs): |
||
| 18 | model = ResNet(Bottleneck, [3, 4, 6, 3], |
||
| 19 | radix=1, cardinality=1, bottleneck_width=64, |
||
| 20 | deep_stem=True, avg_down=True, |
||
| 21 | avd=True, avd_first=True, |
||
| 22 | use_splat=True, dropblock_prob=0.1, |
||
| 23 | name_prefix='resnetv1f_', **kwargs) |
||
| 24 | if pretrained: |
||
| 25 | from .model_store import get_model_file |
||
| 26 | model.load_parameters(get_model_file('resnest50_fast_1s1x64d', |
||
| 27 | root=root), ctx=ctx) |
||
| 28 | return model |
||
| 29 | |||
| 30 | def resnest50_fast_2s1x64d(pretrained=False, root='~/.mxnet/models', ctx=cpu(0), **kwargs): |
||
| 31 | model = ResNet(Bottleneck, [3, 4, 6, 3], |
||
| 32 | radix=2, cardinality=1, bottleneck_width=64, |
||
| 33 | deep_stem=True, avg_down=True, |
||
| 34 | avd=True, avd_first=True, |
||
| 35 | use_splat=True, dropblock_prob=0.1, |
||
| 36 | name_prefix='resnetv1f_', **kwargs) |
||
| 37 | if pretrained: |
||
| 38 | from .model_store import get_model_file |
||
| 39 | model.load_parameters(get_model_file('resnest50_fast_2s1x64d', |
||
| 40 | root=root), ctx=ctx) |
||
| 41 | return model |
||
| 42 | |||
| 43 | def resnest50_fast_4s1x64d(pretrained=False, root='~/.mxnet/models', ctx=cpu(0), **kwargs): |
||
| 44 | model = ResNet(Bottleneck, [3, 4, 6, 3], |
||
| 45 | radix=4, cardinality=1, bottleneck_width=64, |
||
| 46 | deep_stem=True, avg_down=True, |
||
| 47 | avd=True, avd_first=True, |
||
| 48 | use_splat=True, dropblock_prob=0.1, |
||
| 49 | name_prefix='resnetv1f_', **kwargs) |
||
| 50 | if pretrained: |
||
| 51 | from .model_store import get_model_file |
||
| 52 | model.load_parameters(get_model_file('resnest50_fast_4s1x64d', |
||
| 53 | root=root), ctx=ctx) |
||
| 54 | return model |
||
| 55 | |||
| 56 | def resnest50_fast_1s2x40d(pretrained=False, root='~/.mxnet/models', ctx=cpu(0), **kwargs): |
||
| 57 | model = ResNet(Bottleneck, [3, 4, 6, 3], |
||
| 58 | radix=1, cardinality=2, bottleneck_width=40, |
||
| 59 | deep_stem=True, avg_down=True, |
||
| 60 | avd=True, avd_first=True, |
||
| 61 | use_splat=True, dropblock_prob=0.1, |
||
| 62 | name_prefix='resnetv1f_', **kwargs) |
||
| 63 | if pretrained: |
||
| 64 | from .model_store import get_model_file |
||
| 65 | model.load_parameters(get_model_file('resnest50_fast_1s2x40d', |
||
| 66 | root=root), ctx=ctx) |
||
| 67 | return model |
||
| 68 | |||
| 69 | def resnest50_fast_2s2x40d(pretrained=False, root='~/.mxnet/models', ctx=cpu(0), **kwargs): |
||
| 70 | model = ResNet(Bottleneck, [3, 4, 6, 3], |
||
| 71 | radix=2, cardinality=2, bottleneck_width=40, |
||
| 72 | deep_stem=True, avg_down=True, |
||
| 73 | avd=True, avd_first=True, |
||
| 74 | use_splat=True, dropblock_prob=0.1, |
||
| 75 | name_prefix='resnetv1f_', **kwargs) |
||
| 76 | if pretrained: |
||
| 77 | from .model_store import get_model_file |
||
| 78 | model.load_parameters(get_model_file('resnest50_fast_2s2x40d', |
||
| 79 | root=root), ctx=ctx) |
||
| 80 | return model |
||
| 81 | |||
| 82 | def resnest50_fast_4s2x40d(pretrained=False, root='~/.mxnet/models', ctx=cpu(0), **kwargs): |
||
| 83 | model = ResNet(Bottleneck, [3, 4, 6, 3], |
||
| 84 | radix=4, cardinality=2, bottleneck_width=40, |
||
| 85 | deep_stem=True, avg_down=True, |
||
| 86 | avd=True, avd_first=True, |
||
| 87 | use_splat=True, dropblock_prob=0.1, |
||
| 88 | name_prefix='resnetv1f_', **kwargs) |
||
| 89 | if pretrained: |
||
| 90 | from .model_store import get_model_file |
||
| 91 | model.load_parameters(get_model_file('resnest50_fast_4s2x40d', |
||
| 92 | root=root), ctx=ctx) |
||
| 93 | return model |
||
| 94 | |||
| 95 | def resnest50_fast_1s4x24d(pretrained=False, root='~/.mxnet/models', ctx=cpu(0), **kwargs): |
||
| 96 | model = ResNet(Bottleneck, [3, 4, 6, 3], |
||
| 97 | radix=1, cardinality=4, bottleneck_width=24, |
||
| 98 | deep_stem=True, avg_down=True, |
||
| 99 | avd=True, avd_first=True, |
||
| 100 | use_splat=True, dropblock_prob=0.1, |
||
| 101 | name_prefix='resnetv1f_', **kwargs) |
||
| 102 | if pretrained: |
||
| 103 | from .model_store import get_model_file |
||
| 104 | model.load_parameters(get_model_file('resnest50_fast_1s4x24d', |
||
| 105 | root=root), ctx=ctx) |
||
| 106 | return model |
||
| 107 |