weights/state_dict_structure.json

{
    "model": {
        "bert.embeddings.word_embeddings.weight": "Tensor with shape [30528, 1024]",
        "bert.embeddings.position_embeddings.weight": "Tensor with shape [512, 1024]",
        "bert.embeddings.token_type_embeddings.weight": "Tensor with shape [2, 1024]",
        "bert.embeddings.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.embeddings.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.0.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.0.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.0.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.0.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.0.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.0.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.0.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.0.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.0.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.0.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.0.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.0.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.0.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.0.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.0.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.0.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.0.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.0.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.0.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.0.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.0.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.0.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.0.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.0.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.0.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.0.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.0.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.0.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.0.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.0.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.0.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.0.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.0.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.0.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.0.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.0.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.0.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.0.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.0.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.0.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.0.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.0.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.0.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.0.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.0.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.0.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.1.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.1.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.1.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.1.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.1.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.1.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.1.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.1.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.1.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.1.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.1.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.1.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.1.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.1.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.1.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.1.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.1.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.1.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.1.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.1.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.1.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.1.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.1.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.1.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.1.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.1.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.1.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.1.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.1.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.1.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.1.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.1.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.1.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.1.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.1.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.1.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.1.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.1.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.1.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.1.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.1.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.1.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.1.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.1.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.1.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.1.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.2.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.2.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.2.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.2.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.2.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.2.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.2.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.2.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.2.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.2.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.2.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.2.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.2.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.2.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.2.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.2.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.2.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.2.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.2.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.2.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.2.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.2.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.2.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.2.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.2.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.2.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.2.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.2.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.2.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.2.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.2.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.2.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.2.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.2.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.2.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.2.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.2.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.2.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.2.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.2.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.2.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.2.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.2.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.2.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.2.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.2.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.3.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.3.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.3.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.3.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.3.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.3.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.3.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.3.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.3.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.3.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.3.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.3.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.3.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.3.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.3.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.3.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.3.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.3.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.3.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.3.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.3.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.3.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.3.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.3.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.3.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.3.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.3.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.3.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.3.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.3.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.3.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.3.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.3.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.3.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.3.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.3.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.3.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.3.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.3.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.3.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.3.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.3.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.3.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.3.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.3.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.3.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.4.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.4.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.4.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.4.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.4.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.4.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.4.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.4.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.4.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.4.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.4.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.4.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.4.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.4.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.4.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.4.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.4.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.4.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.4.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.4.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.4.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.4.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.4.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.4.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.4.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.4.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.4.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.4.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.4.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.4.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.4.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.4.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.4.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.4.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.4.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.4.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.4.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.4.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.4.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.4.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.4.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.4.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.4.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.4.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.4.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.4.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.5.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.5.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.5.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.5.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.5.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.5.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.5.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.5.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.5.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.5.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.5.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.5.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.5.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.5.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.5.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.5.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.5.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.5.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.5.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.5.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.5.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.5.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.5.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.5.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.5.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.5.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.5.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.5.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.5.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.5.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.5.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.5.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.5.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.5.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.5.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.5.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.5.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.5.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.5.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.5.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.5.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.5.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.5.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.5.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.5.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.5.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.6.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.6.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.6.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.6.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.6.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.6.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.6.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.6.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.6.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.6.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.6.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.6.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.6.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.6.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.6.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.6.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.6.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.6.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.6.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.6.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.6.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.6.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.6.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.6.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.6.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.6.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.6.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.6.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.6.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.6.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.6.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.6.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.6.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.6.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.6.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.6.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.6.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.6.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.6.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.6.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.6.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.6.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.6.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.6.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.6.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.6.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.7.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.7.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.7.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.7.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.7.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.7.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.7.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.7.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.7.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.7.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.7.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.7.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.7.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.7.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.7.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.7.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.7.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.7.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.7.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.7.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.7.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.7.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.7.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.7.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.7.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.7.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.7.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.7.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.7.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.7.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.7.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.7.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.7.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.7.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.7.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.7.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.7.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.7.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.7.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.7.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.7.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.7.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.7.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.7.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.7.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.7.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.8.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.8.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.8.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.8.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.8.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.8.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.8.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.8.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.8.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.8.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.8.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.8.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.8.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.8.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.8.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.8.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.8.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.8.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.8.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.8.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.8.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.8.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.8.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.8.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.8.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.8.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.8.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.8.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.8.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.8.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.8.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.8.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.8.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.8.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.8.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.8.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.8.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.8.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.8.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.8.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.8.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.8.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.8.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.8.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.8.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.8.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.9.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.9.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.9.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.9.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.9.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.9.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.9.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.9.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.9.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.9.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.9.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.9.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.9.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.9.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.9.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.9.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.9.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.9.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.9.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.9.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.9.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.9.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.9.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.9.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.9.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.9.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.9.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.9.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.9.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.9.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.9.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.9.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.9.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.9.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.9.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.9.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.9.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.9.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.9.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.9.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.9.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.9.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.9.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.9.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.9.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.9.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.10.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.10.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.10.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.10.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.10.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.10.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.10.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.10.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.10.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.10.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.10.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.10.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.10.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.10.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.10.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.10.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.10.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.10.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.10.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.10.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.10.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.10.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.10.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.10.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.10.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.10.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.10.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.10.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.10.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.10.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.10.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.10.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.10.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.10.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.10.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.10.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.10.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.10.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.10.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.10.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.10.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.10.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.10.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.10.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.10.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.10.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.11.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.11.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.11.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.11.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.11.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.11.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.11.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.11.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.11.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.11.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.11.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.11.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.11.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.11.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.11.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.11.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.11.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.11.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.11.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.11.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.11.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.11.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.11.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.11.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.11.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.11.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.11.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.11.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.11.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.11.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.11.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.11.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.11.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.11.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.11.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.11.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.11.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.11.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.11.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.11.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.11.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.11.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.11.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.11.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.11.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.11.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.12.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.12.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.12.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.12.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.12.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.12.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.12.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.12.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.12.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.12.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.12.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.12.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.12.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.12.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.12.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.12.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.12.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.12.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.12.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.12.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.12.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.12.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.12.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.12.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.12.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.12.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.12.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.12.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.12.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.12.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.12.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.12.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.12.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.12.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.12.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.12.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.12.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.12.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.12.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.12.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.12.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.12.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.12.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.12.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.12.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.12.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.13.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.13.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.13.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.13.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.13.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.13.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.13.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.13.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.13.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.13.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.13.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.13.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.13.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.13.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.13.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.13.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.13.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.13.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.13.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.13.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.13.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.13.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.13.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.13.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.13.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.13.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.13.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.13.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.13.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.13.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.13.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.13.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.13.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.13.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.13.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.13.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.13.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.13.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.13.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.13.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.13.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.13.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.13.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.13.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.13.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.13.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.14.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.14.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.14.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.14.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.14.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.14.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.14.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.14.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.14.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.14.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.14.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.14.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.14.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.14.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.14.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.14.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.14.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.14.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.14.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.14.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.14.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.14.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.14.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.14.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.14.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.14.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.14.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.14.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.14.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.14.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.14.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.14.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.14.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.14.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.14.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.14.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.14.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.14.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.14.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.14.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.14.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.14.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.14.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.14.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.14.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.14.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.15.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.15.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.15.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.15.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.15.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.15.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.15.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.15.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.15.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.15.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.15.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.15.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.15.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.15.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.15.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.15.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.15.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.15.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.15.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.15.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.15.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.15.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.15.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.15.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.15.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.15.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.15.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.15.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.15.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.15.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.15.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.15.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.15.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.15.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.15.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.15.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.15.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.15.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.15.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.15.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.15.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.15.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.15.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.15.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.15.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.15.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.16.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.16.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.16.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.16.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.16.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.16.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.16.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.16.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.16.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.16.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.16.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.16.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.16.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.16.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.16.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.16.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.16.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.16.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.16.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.16.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.16.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.16.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.16.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.16.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.16.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.16.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.16.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.16.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.16.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.16.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.16.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.16.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.16.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.16.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.16.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.16.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.16.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.16.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.16.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.16.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.16.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.16.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.16.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.16.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.16.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.16.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.17.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.17.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.17.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.17.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.17.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.17.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.17.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.17.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.17.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.17.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.17.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.17.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.17.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.17.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.17.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.17.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.17.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.17.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.17.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.17.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.17.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.17.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.17.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.17.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.17.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.17.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.17.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.17.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.17.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.17.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.17.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.17.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.17.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.17.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.17.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.17.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.17.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.17.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.17.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.17.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.17.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.17.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.17.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.17.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.17.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.17.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.18.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.18.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.18.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.18.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.18.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.18.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.18.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.18.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.18.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.18.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.18.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.18.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.18.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.18.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.18.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.18.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.18.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.18.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.18.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.18.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.18.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.18.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.18.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.18.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.18.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.18.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.18.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.18.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.18.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.18.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.18.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.18.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.18.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.18.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.18.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.18.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.18.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.18.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.18.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.18.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.18.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.18.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.18.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.18.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.18.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.18.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.19.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.19.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.19.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.19.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.19.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.19.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.19.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.19.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.19.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.19.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.19.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.19.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.19.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.19.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.19.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.19.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.19.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.19.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.19.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.19.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.19.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.19.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.19.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.19.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.19.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.19.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.19.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.19.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.19.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.19.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.19.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.19.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.19.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.19.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.19.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.19.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.19.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.19.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.19.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.19.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.19.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.19.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.19.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.19.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.19.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.19.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.20.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.20.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.20.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.20.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.20.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.20.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.20.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.20.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.20.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.20.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.20.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.20.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.20.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.20.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.20.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.20.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.20.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.20.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.20.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.20.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.20.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.20.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.20.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.20.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.20.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.20.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.20.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.20.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.20.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.20.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.20.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.20.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.20.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.20.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.20.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.20.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.20.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.20.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.20.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.20.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.20.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.20.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.20.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.20.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.20.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.20.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.21.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.21.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.21.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.21.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.21.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.21.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.21.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.21.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.21.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.21.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.21.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.21.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.21.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.21.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.21.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.21.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.21.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.21.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.21.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.21.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.21.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.21.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.21.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.21.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.21.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.21.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.21.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.21.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.21.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.21.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.21.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.21.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.21.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.21.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.21.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.21.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.21.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.21.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.21.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.21.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.21.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.21.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.21.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.21.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.21.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.21.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.22.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.22.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.22.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.22.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.22.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.22.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.22.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.22.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.22.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.22.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.22.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.22.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.22.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.22.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.22.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.22.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.22.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.22.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.22.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.22.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.22.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.22.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.22.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.22.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.22.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.22.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.22.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.22.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.22.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.22.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.22.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.22.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.22.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.22.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.22.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.22.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.22.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.22.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.22.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.22.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.22.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.22.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.22.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.22.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.22.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.22.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.23.attention.self.query.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.23.attention.self.query.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.23.attention.self.query.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.23.attention.self.query.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.23.attention.self.query.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.23.attention.self.query.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.23.attention.self.query.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.23.attention.self.key.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.23.attention.self.key.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.23.attention.self.key.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.23.attention.self.key.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.23.attention.self.key.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.23.attention.self.key.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.23.attention.self.key.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.23.attention.self.value.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.23.attention.self.value.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.23.attention.self.value.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.23.attention.self.value.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.23.attention.self.value.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.23.attention.self.value.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.23.attention.self.value.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.23.attention.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.23.attention.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.23.attention.output.dense.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.23.attention.output.dense.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.23.attention.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.23.attention.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.23.attention.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.23.attention.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.23.attention.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.23.intermediate.dense_act.linear.bias": "Tensor with shape [4096]",
        "bert.encoder.layer.23.intermediate.dense_act.linear.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.23.intermediate.dense_act.linear.tensor.factors.1": "Tensor with shape [12, 8, 72]",
        "bert.encoder.layer.23.intermediate.dense_act.linear.tensor.factors.2": "Tensor with shape [72, 8, 300]",
        "bert.encoder.layer.23.intermediate.dense_act.linear.tensor.factors.3": "Tensor with shape [300, 16, 192]",
        "bert.encoder.layer.23.intermediate.dense_act.linear.tensor.factors.4": "Tensor with shape [192, 16, 12]",
        "bert.encoder.layer.23.intermediate.dense_act.linear.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.23.output.dense.bias": "Tensor with shape [1024]",
        "bert.encoder.layer.23.output.dense.tensor.factors.0": "Tensor with shape [1, 16, 12]",
        "bert.encoder.layer.23.output.dense.tensor.factors.1": "Tensor with shape [12, 16, 144]",
        "bert.encoder.layer.23.output.dense.tensor.factors.2": "Tensor with shape [144, 16, 300]",
        "bert.encoder.layer.23.output.dense.tensor.factors.3": "Tensor with shape [300, 8, 96]",
        "bert.encoder.layer.23.output.dense.tensor.factors.4": "Tensor with shape [96, 8, 12]",
        "bert.encoder.layer.23.output.dense.tensor.factors.5": "Tensor with shape [12, 16, 1]",
        "bert.encoder.layer.23.output.LayerNorm.weight": "Tensor with shape [1024]",
        "bert.encoder.layer.23.output.LayerNorm.bias": "Tensor with shape [1024]",
        "bert.pooler.dense_act.linear.weight": "Tensor with shape [1024, 1024]",
        "bert.pooler.dense_act.linear.bias": "Tensor with shape [1024]",
        "cls.predictions.bias": "Tensor with shape [30528]",
        "cls.predictions.transform.dense_act.linear.weight": "Tensor with shape [1024, 1024]",
        "cls.predictions.transform.dense_act.linear.bias": "Tensor with shape [1024]",
        "cls.predictions.transform.LayerNorm.weight": "Tensor with shape [1024]",
        "cls.predictions.transform.LayerNorm.bias": "Tensor with shape [1024]",
        "cls.predictions.decoder.weight": "Tensor with shape [30528, 1024]",
        "cls.seq_relationship.weight": "Tensor with shape [2, 1024]",
        "cls.seq_relationship.bias": "Tensor with shape [2]"
    },
    "optimizer": {
        "state": {
            "0": {
                "exp_avg": "Tensor with shape [30528, 1024]",
                "exp_avg_sq": "Tensor with shape [30528, 1024]"
            },
            "1": {
                "exp_avg": "Tensor with shape [512, 1024]",
                "exp_avg_sq": "Tensor with shape [512, 1024]"
            },
            "2": {
                "exp_avg": "Tensor with shape [2, 1024]",
                "exp_avg_sq": "Tensor with shape [2, 1024]"
            },
            "3": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "4": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "5": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "6": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "7": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "8": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "9": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "10": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "11": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "12": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "13": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "14": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "15": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "16": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "17": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "18": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "19": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "20": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "21": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "22": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "23": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "24": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "25": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "26": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "27": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "28": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "29": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "30": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "31": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "32": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "33": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "34": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "35": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "36": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "37": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "38": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "39": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "40": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "41": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "42": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "43": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "44": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "45": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "46": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "47": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "48": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "49": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "50": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "51": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "52": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "53": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "54": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "55": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "56": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "57": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "58": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "59": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "60": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "61": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "62": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "63": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "64": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "65": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "66": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "67": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "68": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "69": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "70": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "71": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "72": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "73": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "74": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "75": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "76": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "77": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "78": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "79": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "80": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "81": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "82": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "83": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "84": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "85": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "86": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "87": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "88": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "89": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "90": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "91": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "92": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "93": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "94": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "95": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "96": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "97": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "98": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "99": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "100": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "101": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "102": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "103": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "104": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "105": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "106": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "107": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "108": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "109": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "110": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "111": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "112": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "113": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "114": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "115": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "116": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "117": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "118": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "119": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "120": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "121": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "122": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "123": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "124": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "125": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "126": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "127": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "128": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "129": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "130": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "131": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "132": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "133": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "134": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "135": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "136": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "137": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "138": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "139": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "140": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "141": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "142": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "143": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "144": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "145": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "146": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "147": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "148": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "149": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "150": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "151": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "152": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "153": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "154": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "155": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "156": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "157": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "158": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "159": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "160": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "161": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "162": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "163": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "164": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "165": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "166": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "167": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "168": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "169": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "170": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "171": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "172": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "173": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "174": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "175": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "176": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "177": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "178": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "179": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "180": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "181": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "182": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "183": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "184": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "185": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "186": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "187": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "188": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "189": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "190": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "191": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "192": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "193": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "194": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "195": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "196": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "197": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "198": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "199": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "200": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "201": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "202": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "203": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "204": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "205": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "206": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "207": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "208": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "209": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "210": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "211": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "212": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "213": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "214": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "215": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "216": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "217": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "218": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "219": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "220": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "221": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "222": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "223": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "224": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "225": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "226": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "227": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "228": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "229": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "230": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "231": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "232": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "233": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "234": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "235": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "236": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "237": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "238": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "239": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "240": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "241": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "242": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "243": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "244": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "245": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "246": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "247": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "248": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "249": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "250": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "251": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "252": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "253": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "254": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "255": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "256": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "257": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "258": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "259": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "260": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "261": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "262": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "263": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "264": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "265": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "266": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "267": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "268": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "269": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "270": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "271": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "272": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "273": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "274": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "275": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "276": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "277": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "278": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "279": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "280": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "281": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "282": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "283": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "284": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "285": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "286": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "287": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "288": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "289": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "290": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "291": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "292": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "293": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "294": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "295": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "296": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "297": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "298": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "299": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "300": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "301": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "302": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "303": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "304": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "305": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "306": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "307": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "308": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "309": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "310": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "311": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "312": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "313": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "314": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "315": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "316": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "317": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "318": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "319": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "320": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "321": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "322": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "323": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "324": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "325": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "326": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "327": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "328": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "329": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "330": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "331": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "332": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "333": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "334": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "335": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "336": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "337": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "338": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "339": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "340": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "341": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "342": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "343": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "344": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "345": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "346": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "347": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "348": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "349": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "350": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "351": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "352": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "353": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "354": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "355": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "356": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "357": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "358": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "359": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "360": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "361": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "362": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "363": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "364": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "365": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "366": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "367": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "368": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "369": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "370": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "371": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "372": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "373": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "374": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "375": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "376": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "377": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "378": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "379": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "380": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "381": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "382": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "383": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "384": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "385": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "386": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "387": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "388": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "389": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "390": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "391": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "392": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "393": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "394": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "395": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "396": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "397": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "398": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "399": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "400": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "401": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "402": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "403": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "404": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "405": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "406": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "407": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "408": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "409": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "410": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "411": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "412": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "413": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "414": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "415": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "416": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "417": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "418": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "419": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "420": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "421": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "422": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "423": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "424": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "425": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "426": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "427": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "428": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "429": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "430": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "431": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "432": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "433": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "434": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "435": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "436": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "437": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "438": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "439": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "440": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "441": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "442": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "443": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "444": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "445": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "446": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "447": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "448": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "449": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "450": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "451": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "452": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "453": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "454": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "455": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "456": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "457": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "458": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "459": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "460": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "461": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "462": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "463": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "464": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "465": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "466": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "467": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "468": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "469": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "470": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "471": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "472": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "473": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "474": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "475": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "476": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "477": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "478": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "479": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "480": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "481": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "482": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "483": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "484": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "485": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "486": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "487": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "488": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "489": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "490": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "491": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "492": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "493": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "494": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "495": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "496": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "497": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "498": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "499": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "500": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "501": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "502": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "503": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "504": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "505": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "506": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "507": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "508": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "509": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "510": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "511": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "512": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "513": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "514": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "515": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "516": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "517": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "518": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "519": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "520": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "521": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "522": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "523": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "524": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "525": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "526": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "527": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "528": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "529": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "530": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "531": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "532": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "533": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "534": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "535": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "536": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "537": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "538": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "539": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "540": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "541": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "542": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "543": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "544": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "545": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "546": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "547": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "548": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "549": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "550": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "551": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "552": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "553": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "554": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "555": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "556": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "557": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "558": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "559": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "560": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "561": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "562": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "563": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "564": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "565": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "566": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "567": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "568": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "569": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "570": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "571": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "572": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "573": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "574": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "575": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "576": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "577": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "578": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "579": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "580": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "581": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "582": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "583": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "584": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "585": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "586": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "587": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "588": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "589": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "590": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "591": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "592": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "593": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "594": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "595": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "596": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "597": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "598": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "599": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "600": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "601": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "602": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "603": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "604": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "605": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "606": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "607": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "608": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "609": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "610": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "611": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "612": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "613": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "614": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "615": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "616": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "617": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "618": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "619": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "620": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "621": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "622": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "623": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "624": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "625": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "626": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "627": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "628": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "629": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "630": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "631": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "632": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "633": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "634": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "635": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "636": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "637": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "638": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "639": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "640": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "641": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "642": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "643": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "644": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "645": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "646": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "647": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "648": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "649": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "650": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "651": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "652": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "653": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "654": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "655": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "656": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "657": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "658": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "659": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "660": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "661": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "662": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "663": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "664": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "665": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "666": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "667": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "668": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "669": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "670": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "671": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "672": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "673": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "674": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "675": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "676": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "677": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "678": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "679": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "680": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "681": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "682": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "683": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "684": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "685": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "686": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "687": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "688": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "689": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "690": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "691": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "692": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "693": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "694": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "695": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "696": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "697": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "698": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "699": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "700": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "701": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "702": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "703": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "704": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "705": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "706": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "707": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "708": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "709": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "710": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "711": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "712": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "713": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "714": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "715": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "716": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "717": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "718": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "719": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "720": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "721": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "722": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "723": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "724": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "725": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "726": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "727": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "728": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "729": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "730": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "731": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "732": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "733": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "734": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "735": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "736": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "737": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "738": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "739": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "740": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "741": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "742": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "743": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "744": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "745": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "746": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "747": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "748": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "749": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "750": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "751": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "752": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "753": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "754": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "755": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "756": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "757": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "758": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "759": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "760": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "761": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "762": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "763": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "764": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "765": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "766": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "767": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "768": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "769": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "770": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "771": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "772": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "773": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "774": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "775": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "776": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "777": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "778": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "779": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "780": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "781": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "782": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "783": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "784": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "785": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "786": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "787": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "788": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "789": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "790": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "791": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "792": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "793": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "794": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "795": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "796": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "797": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "798": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "799": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "800": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "801": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "802": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "803": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "804": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "805": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "806": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "807": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "808": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "809": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "810": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "811": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "812": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "813": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "814": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "815": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "816": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "817": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "818": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "819": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "820": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "821": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "822": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "823": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "824": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "825": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "826": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "827": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "828": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "829": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "830": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "831": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "832": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "833": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "834": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "835": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "836": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "837": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "838": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "839": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "840": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "841": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "842": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "843": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "844": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "845": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "846": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "847": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "848": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "849": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "850": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "851": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "852": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "853": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "854": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "855": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "856": {
                "exp_avg": "Tensor with shape [12, 8, 72]",
                "exp_avg_sq": "Tensor with shape [12, 8, 72]"
            },
            "857": {
                "exp_avg": "Tensor with shape [72, 8, 300]",
                "exp_avg_sq": "Tensor with shape [72, 8, 300]"
            },
            "858": {
                "exp_avg": "Tensor with shape [300, 16, 192]",
                "exp_avg_sq": "Tensor with shape [300, 16, 192]"
            },
            "859": {
                "exp_avg": "Tensor with shape [192, 16, 12]",
                "exp_avg_sq": "Tensor with shape [192, 16, 12]"
            },
            "860": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "861": {
                "exp_avg": "Tensor with shape [1, 16, 12]",
                "exp_avg_sq": "Tensor with shape [1, 16, 12]"
            },
            "862": {
                "exp_avg": "Tensor with shape [12, 16, 144]",
                "exp_avg_sq": "Tensor with shape [12, 16, 144]"
            },
            "863": {
                "exp_avg": "Tensor with shape [144, 16, 300]",
                "exp_avg_sq": "Tensor with shape [144, 16, 300]"
            },
            "864": {
                "exp_avg": "Tensor with shape [300, 8, 96]",
                "exp_avg_sq": "Tensor with shape [300, 8, 96]"
            },
            "865": {
                "exp_avg": "Tensor with shape [96, 8, 12]",
                "exp_avg_sq": "Tensor with shape [96, 8, 12]"
            },
            "866": {
                "exp_avg": "Tensor with shape [12, 16, 1]",
                "exp_avg_sq": "Tensor with shape [12, 16, 1]"
            },
            "867": {
                "exp_avg": "Tensor with shape [1024, 1024]",
                "exp_avg_sq": "Tensor with shape [1024, 1024]"
            },
            "868": {
                "exp_avg": "Tensor with shape [1024, 1024]",
                "exp_avg_sq": "Tensor with shape [1024, 1024]"
            },
            "869": {
                "exp_avg": "Tensor with shape [2, 1024]",
                "exp_avg_sq": "Tensor with shape [2, 1024]"
            },
            "870": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "871": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "872": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "873": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "874": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "875": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "876": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "877": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "878": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "879": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "880": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "881": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "882": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "883": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "884": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "885": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "886": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "887": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "888": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "889": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "890": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "891": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "892": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "893": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "894": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "895": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "896": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "897": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "898": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "899": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "900": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "901": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "902": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "903": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "904": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "905": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "906": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "907": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "908": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "909": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "910": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "911": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "912": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "913": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "914": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "915": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "916": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "917": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "918": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "919": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "920": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "921": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "922": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "923": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "924": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "925": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "926": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "927": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "928": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "929": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "930": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "931": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "932": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "933": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "934": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "935": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "936": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "937": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "938": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "939": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "940": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "941": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "942": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "943": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "944": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "945": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "946": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "947": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "948": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "949": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "950": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "951": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "952": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "953": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "954": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "955": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "956": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "957": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "958": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "959": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "960": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "961": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "962": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "963": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "964": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "965": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "966": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "967": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "968": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "969": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "970": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "971": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "972": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "973": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "974": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "975": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "976": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "977": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "978": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "979": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "980": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "981": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "982": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "983": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "984": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "985": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "986": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "987": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "988": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "989": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "990": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "991": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "992": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "993": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "994": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "995": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "996": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "997": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "998": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "999": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1000": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1001": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1002": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1003": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1004": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1005": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1006": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1007": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1008": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1009": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1010": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1011": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1012": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1013": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1014": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1015": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1016": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1017": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1018": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1019": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1020": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1021": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1022": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1023": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1024": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1025": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1026": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1027": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1028": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1029": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1030": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1031": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1032": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1033": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1034": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1035": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1036": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1037": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1038": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1039": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1040": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1041": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1042": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1043": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1044": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1045": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1046": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1047": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1048": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1049": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1050": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1051": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1052": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1053": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1054": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1055": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1056": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1057": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1058": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1059": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1060": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1061": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1062": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1063": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1064": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1065": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1066": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1067": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1068": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1069": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1070": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1071": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1072": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1073": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1074": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1075": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1076": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1077": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1078": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1079": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1080": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1081": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1082": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1083": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1084": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1085": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1086": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1087": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1088": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1089": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1090": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1091": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1092": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1093": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1094": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1095": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1096": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1097": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1098": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1099": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1100": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1101": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1102": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1103": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1104": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1105": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1106": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1107": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1108": {
                "exp_avg": "Tensor with shape [4096]",
                "exp_avg_sq": "Tensor with shape [4096]"
            },
            "1109": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1110": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1111": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1112": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1113": {
                "exp_avg": "Tensor with shape [30528]",
                "exp_avg_sq": "Tensor with shape [30528]"
            },
            "1114": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1115": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1116": {
                "exp_avg": "Tensor with shape [1024]",
                "exp_avg_sq": "Tensor with shape [1024]"
            },
            "1117": {
                "exp_avg": "Tensor with shape [2]",
                "exp_avg_sq": "Tensor with shape [2]"
            }
        },
        "param_groups": "<class 'list'>"
    },
    "grad_scaler": {
        "scale": 2048.0,
        "growth_factor": 2.0,
        "backoff_factor": 0.5,
        "growth_interval": 2000,
        "_growth_tracker": 1563
    },
    "epoch": 2
}