Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- DetectionModel(
- (model): Sequential(
- (0): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 3, 16, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 16, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (2): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 32, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 48, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 16, 8, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 8, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- (3): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (4): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 96, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- (5): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (6): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 192, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): C3k(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 32, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 32, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv3): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): Sequential(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (1): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- )
- )
- (7): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (8): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 384, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): C3k(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv3): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): Sequential(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (1): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- )
- )
- (9): SPPF(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 512, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): MaxPool2d(kernel_size=5, stride=1, padding=2, dilation=1, ceil_mode=False)
- )
- (10): C2PSA(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): Sequential(
- (0): PSABlock(
- (attn): Attention(
- (qkv): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): Conv2d(
- 128, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (proj): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): Conv2d(
- 128, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (pe): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): Conv2d(
- 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (ffn): Sequential(
- (0): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): Conv2d(
- 256, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- )
- (11): Upsample(scale_factor=2.0, mode='nearest')
- (12): Concat()
- (13): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 384, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 192, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- (14): Upsample(scale_factor=2.0, mode='nearest')
- (15): Concat()
- (16): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 96, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- (17): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (18): Concat()
- (19): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 192, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 192, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- (20): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (21): Concat()
- (22): C3k2(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 384, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 384, 256, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): ModuleList(
- (0): C3k(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv3): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 128, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (m): Sequential(
- (0): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (1): Bottleneck(
- (cv1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (cv2): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- )
- )
- )
- )
- (23): Detect(
- (cv2): ModuleList(
- (0): Sequential(
- (0): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (2): Conv2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- )
- (1): Sequential(
- (0): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (2): Conv2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- )
- (2): Sequential(
- (0): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (2): Conv2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- )
- )
- (cv3): ModuleList(
- (0): Sequential(
- (0): Sequential(
- (0): DWConv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (1): Sequential(
- (0): DWConv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (2): Conv2d(
- 64, 2, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- )
- (1): Sequential(
- (0): Sequential(
- (0): DWConv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 128, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (1): Sequential(
- (0): DWConv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (2): Conv2d(
- 64, 2, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- )
- (2): Sequential(
- (0): Sequential(
- (0): DWConv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 256, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (1): Sequential(
- (0): DWConv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=64
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- (1): Conv(
- (quant): QuantStub(
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (dequant): DeQuantStub()
- (conv): ConvReLU2d(
- 64, 64, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- (bn): Identity()
- (act): Identity()
- )
- )
- (2): Conv2d(
- 64, 2, kernel_size=(1, 1), stride=(1, 1)
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- )
- )
- (dfl): DFL(
- (conv): Conv2d(
- 16, 1, kernel_size=(1, 1), stride=(1, 1), bias=False
- (weight_fake_quant): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.qint8, quant_min=-128, quant_max=127, qscheme=torch.per_channel_symmetric, reduce_range=False
- (activation_post_process): MovingAveragePerChannelMinMaxObserver(min_val=tensor([], device='cuda:0'), max_val=tensor([], device='cuda:0'))
- )
- (activation_post_process): FusedMovingAvgObsFakeQuantize(
- fake_quant_enabled=tensor([1], device='cuda:0'), observer_enabled=tensor([1], device='cuda:0'), scale=tensor([1.], device='cuda:0'), zero_point=tensor([0], device='cuda:0', dtype=torch.int32), dtype=torch.quint8, quant_min=0, quant_max=127, qscheme=torch.per_tensor_affine, reduce_range=True
- (activation_post_process): MovingAverageMinMaxObserver(min_val=inf, max_val=-inf)
- )
- )
- )
- )
- )
- )
Advertisement
Add Comment
Please, Sign In to add comment