fastnas

Module implementing fasnas pruning algorithm for search.

Classes

`BinarySearcher`	An iterative searcher that uses binary search to find the best configuration.
`FastNASModeDescriptor`	Class to describe the `"fastnas"` mode.
`FastNASPatchManager`	A patch manager for FastNAS (same as AutoNAS except no sampling during training).

Functions

`convert_fastnas_searchspace`	Convert search space for FastNAS mode with correct patch manager.
`restore_fastnas_searchspace`	Restore search space for FastNAS mode with correct patch manager.

class BinarySearcher

Bases: IterativeSearcher

An iterative searcher that uses binary search to find the best configuration.

after_step()

Update boundaries of the interval after recursing.

Return type:: None

before_search()

Build sensitivity map before search that we use to approximate the cost function.

Return type:: None

before_step()

Check what the middle value is to determine where we recurse.

Return type:: None

property default_state_dict: dict[str, Any]: We also store the sensitivity map and related arguments.

early_stop()

Early stop if the interval is small enough.

Return type:: bool

property hparam_names_for_search: set[str]: We can only optimize over certain types of hparams in binary search.

property hparam_types_for_search: tuple[type]: We can only optimize over certain types of hparams in binary search.

load_search_checkpoint()

We only want to load sensitivity map and original_score here and keep the rest.

Return type:: bool

max_degrade: float

middle_value: float

min_degrade: float

original_score: float

sample()

Check in which interval we should recurse and sets the corresponding subnet.

Return type:: dict[str, Any]

sensitivity_map: dict[str, dict[int, float]]

ModeloptConfig FastNASConfig

Bases: ModeloptBaseRuleConfig

Configuration for the "fastnas" mode.

Show default config as JSON

Default config (JSON):

{
   "nn.Conv1d": {
      "*": {
         "channel_divisor": 32,
         "channels_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ],
         "kernel_size": []
      }
   },
   "nn.Conv2d": {
      "*": {
         "channel_divisor": 32,
         "channels_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ],
         "kernel_size": []
      }
   },
   "nn.Conv3d": {
      "*": {
         "channel_divisor": 32,
         "channels_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ],
         "kernel_size": []
      }
   },
   "nn.ConvTranspose1d": {
      "*": {
         "channel_divisor": 32,
         "channels_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ],
         "kernel_size": []
      }
   },
   "nn.ConvTranspose2d": {
      "*": {
         "channel_divisor": 32,
         "channels_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ],
         "kernel_size": []
      }
   },
   "nn.ConvTranspose3d": {
      "*": {
         "channel_divisor": 32,
         "channels_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ],
         "kernel_size": []
      }
   },
   "nn.Linear": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.BatchNorm1d": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.BatchNorm2d": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.BatchNorm3d": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.SyncBatchNorm": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.InstanceNorm1d": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.InstanceNorm2d": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.InstanceNorm3d": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.LayerNorm": {
      "*": {
         "feature_divisor": 32,
         "features_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   },
   "nn.GroupNorm": {
      "*": {
         "channel_divisor": 32,
         "channels_ratio": [
            0.05,
            0.1,
            0.15000000000000002,
            0.2,
            0.25,
            0.30000000000000004,
            0.35000000000000003,
            0.4,
            0.45,
            0.5,
            0.55,
            0.6000000000000001,
            0.65,
            0.7000000000000001,
            0.75,
            0.8,
            0.8500000000000001,
            0.9,
            0.9500000000000001,
            1.0
         ]
      }
   }
}

field nn.BatchNorm1d: DynamicBatchNorm1dConfig | None | dict[str, DynamicBatchNorm1dConfig | None]

Show details

Configuration for dynamic nn.BatchNorm1d module.

If the "nn.BatchNorm1d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.BatchNorm1d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.BatchNorm1d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.BatchNorm2d: DynamicBatchNorm2dConfig | None | dict[str, DynamicBatchNorm2dConfig | None]

Show details

Configuration for dynamic nn.BatchNorm2d module.

If the "nn.BatchNorm2d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.BatchNorm2d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.BatchNorm2d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.BatchNorm3d: DynamicBatchNorm3dConfig | None | dict[str, DynamicBatchNorm3dConfig | None]

Show details

Configuration for dynamic nn.BatchNorm3d module.

If the "nn.BatchNorm3d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.BatchNorm3d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.BatchNorm3d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.Conv1d: DynamicConv1dConfig | None | dict[str, DynamicConv1dConfig | None]

Show details

Configuration for dynamic nn.Conv1d module.

If the "nn.Conv1d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "channels_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "kernel_size": [],
    "channel_divisor": 32
  }
}

To deactivate any dynamic nn.Conv1d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.Conv1d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.Conv2d: DynamicConv2dConfig | None | dict[str, DynamicConv2dConfig | None]

Show details

Configuration for dynamic nn.Conv2d module.

If the "nn.Conv2d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "channels_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "kernel_size": [],
    "channel_divisor": 32
  }
}

To deactivate any dynamic nn.Conv2d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.Conv2d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.Conv3d: DynamicConv3dConfig | None | dict[str, DynamicConv3dConfig | None]

Show details

Configuration for dynamic nn.Conv3d module.

If the "nn.Conv3d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "channels_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "kernel_size": [],
    "channel_divisor": 32
  }
}

To deactivate any dynamic nn.Conv3d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.Conv3d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.ConvTranspose1d: DynamicConvTranspose1dConfig | None | dict[str, DynamicConvTranspose1dConfig | None]

Show details

Configuration for dynamic nn.ConvTranspose1d module.

If the "nn.ConvTranspose1d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "channels_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "kernel_size": [],
    "channel_divisor": 32
  }
}

To deactivate any dynamic nn.ConvTranspose1d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.ConvTranspose1d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.ConvTranspose2d: DynamicConvTranspose2dConfig | None | dict[str, DynamicConvTranspose2dConfig | None]

Show details

Configuration for dynamic nn.ConvTranspose2d module.

If the "nn.ConvTranspose2d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "channels_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "kernel_size": [],
    "channel_divisor": 32
  }
}

To deactivate any dynamic nn.ConvTranspose2d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.ConvTranspose2d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.ConvTranspose3d: DynamicConvTranspose3dConfig | None | dict[str, DynamicConvTranspose3dConfig | None]

Show details

Configuration for dynamic nn.ConvTranspose3d module.

If the "nn.ConvTranspose3d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "channels_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "kernel_size": [],
    "channel_divisor": 32
  }
}

To deactivate any dynamic nn.ConvTranspose3d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.ConvTranspose3d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.GroupNorm: DynamicGroupNormConfig | None | dict[str, DynamicGroupNormConfig | None]

Show details

Configuration for dynamic nn.GroupNorm module.

If the "nn.GroupNorm" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "channels_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "channel_divisor": 32
  }
}

To deactivate any dynamic nn.GroupNorm module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.GroupNorm layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.InstanceNorm1d: DynamicInstanceNorm1dConfig | None | dict[str, DynamicInstanceNorm1dConfig | None]

Show details

Configuration for dynamic nn.InstanceNorm1d module.

If the "nn.InstanceNorm1d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.InstanceNorm1d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.InstanceNorm1d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.InstanceNorm2d: DynamicInstanceNorm2dConfig | None | dict[str, DynamicInstanceNorm2dConfig | None]

Show details

Configuration for dynamic nn.InstanceNorm2d module.

If the "nn.InstanceNorm2d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.InstanceNorm2d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.InstanceNorm2d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.InstanceNorm3d: DynamicInstanceNorm3dConfig | None | dict[str, DynamicInstanceNorm3dConfig | None]

Show details

Configuration for dynamic nn.InstanceNorm3d module.

If the "nn.InstanceNorm3d" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.InstanceNorm3d module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.InstanceNorm3d layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.LayerNorm: DynamicLayerNormConfig | None | dict[str, DynamicLayerNormConfig | None]

Show details

Configuration for dynamic nn.LayerNorm module.

If the "nn.LayerNorm" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.LayerNorm module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.LayerNorm layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.Linear: DynamicLinearConfig | None | dict[str, DynamicLinearConfig | None]

Show details

Configuration for dynamic nn.Linear module.

If the "nn.Linear" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.Linear module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.Linear layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

field nn.SyncBatchNorm: DynamicSyncBatchNormConfig | None | dict[str, DynamicSyncBatchNormConfig | None]

Show details

Configuration for dynamic nn.SyncBatchNorm module.

If the "nn.SyncBatchNorm" key is not specified, the default configuration (shown in JSON) will be used:

{
  "*": {
    "features_ratio": [
05,
1,
15000000000000002,
2,
25,
30000000000000004,
35000000000000003,
4,
45,
5,
55,
6000000000000001,
65,
7000000000000001,
75,
8,
8500000000000001,
9,
9500000000000001,
0
    ],
    "feature_divisor": 32
  }
}

To deactivate any dynamic nn.SyncBatchNorm module, use None instead of providing a dictionary {}.

To specify layer-specific configurations, you can specify a config for each submodule with the key specifying a glob pattern that matches the submodule name. For example, to convert to a dynamic module for all nn.SyncBatchNorm layers except for those in the "lm_head" submodule use:

{
    "*": {...},
    "*lm_head*": None,
}

Note that glob expressions are processed sequentially in the order they are specified. Later keys in the config will overwrite earlier keys if they match the same submodule name.

If you want to specify the same configuration for all submodules, you can provide an unnested dictionary as well:

{...}

which is short for

{
    "*": {...},
}

class FastNASModeDescriptor

Bases: ModeDescriptor

Class to describe the "fastnas" mode.

The properties of this mode can be inspected via the source code.

property config_class: type[ModeloptBaseConfig]: Specifies the config class for the mode.

property convert: Callable[[Module, ModeloptBaseConfig], tuple[Module, dict[str, Any]]] | Callable[[Module, ModeloptBaseConfig, Any], tuple[Module, dict[str, Any]]]: The mode’s entrypoint for converting a model.

property export_mode: str | None: The mode that corresponds to the export mode of this mode.

property name: str: Returns the value (str representation) of the mode.

property next_modes: set[str] | None: Modes that must immediately follow this mode.

property restore: Callable[[Module, ModeloptBaseConfig, dict[str, Any]], Module]: The mode’s entrypoint for restoring a model.

property search_algorithm: type[BaseSearcher]: Specifies the search algorithm to use for this mode (if any).

property update_for_new_mode: Callable[[Module, ModeloptBaseConfig, dict[str, Any]], None]: The mode’s entrypoint for updating the models state before new mode.

property update_for_save: Callable[[Module, ModeloptBaseConfig, dict[str, Any]], None]: The mode’s entrypoint for updating the models state before saving.

class FastNASPatchManager

Bases: AutoNASPatchManager

A patch manager for FastNAS (same as AutoNAS except no sampling during training).

property sample_during_training: Indicates whether we should sample a new subnet during training.

convert_fastnas_searchspace(model, config)

Convert search space for FastNAS mode with correct patch manager.

Parameters:

model (Module)
config (ModeloptBaseConfig)

Return type:

tuple[Module, dict[str, Any]]

restore_fastnas_searchspace(model, config, metadata)

Restore search space for FastNAS mode with correct patch manager.

Parameters:

model (Module)
config (ModeloptBaseConfig)
metadata (dict[str, Any])

Return type:

Module