{ "style": { "offset": 0, "count": 32, "shape": [ 1, 32 ] }, "bert.embeddings.word_embeddings.weight": { "offset": 32, "count": 22784, "shape": [ 178, 128 ] }, "predictor.text_encoder.lstms.0.weight_ih_l0": { "offset": 22816, "count": 86016, "shape": [ 384, 224 ] }, "predictor.text_encoder.lstms.0.weight_hh_l0": { "offset": 108832, "count": 36864, "shape": [ 384, 96 ] }, "predictor.text_encoder.lstms.0.bias_ih_l0": { "offset": 145696, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.0.bias_hh_l0": { "offset": 146080, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.0.weight_ih_l0_reverse": { "offset": 146464, "count": 86016, "shape": [ 384, 224 ] }, "predictor.text_encoder.lstms.0.weight_hh_l0_reverse": { "offset": 232480, "count": 36864, "shape": [ 384, 96 ] }, "predictor.text_encoder.lstms.0.bias_ih_l0_reverse": { "offset": 269344, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.0.bias_hh_l0_reverse": { "offset": 269728, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.1.fc.weight": { "offset": 270112, "count": 12288, "shape": [ 384, 32 ] }, "predictor.text_encoder.lstms.1.fc.bias": { "offset": 282400, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.2.weight_ih_l0": { "offset": 282784, "count": 86016, "shape": [ 384, 224 ] }, "predictor.text_encoder.lstms.2.weight_hh_l0": { "offset": 368800, "count": 36864, "shape": [ 384, 96 ] }, "predictor.text_encoder.lstms.2.bias_ih_l0": { "offset": 405664, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.2.bias_hh_l0": { "offset": 406048, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.2.weight_ih_l0_reverse": { "offset": 406432, "count": 86016, "shape": [ 384, 224 ] }, "predictor.text_encoder.lstms.2.weight_hh_l0_reverse": { "offset": 492448, "count": 36864, "shape": [ 384, 96 ] }, "predictor.text_encoder.lstms.2.bias_ih_l0_reverse": { "offset": 529312, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.2.bias_hh_l0_reverse": { "offset": 529696, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.3.fc.weight": { "offset": 530080, "count": 12288, "shape": [ 384, 32 ] }, "predictor.text_encoder.lstms.3.fc.bias": { "offset": 542368, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.4.weight_ih_l0": { "offset": 542752, "count": 86016, "shape": [ 384, 224 ] }, "predictor.text_encoder.lstms.4.weight_hh_l0": { "offset": 628768, "count": 36864, "shape": [ 384, 96 ] }, "predictor.text_encoder.lstms.4.bias_ih_l0": { "offset": 665632, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.4.bias_hh_l0": { "offset": 666016, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.4.weight_ih_l0_reverse": { "offset": 666400, "count": 86016, "shape": [ 384, 224 ] }, "predictor.text_encoder.lstms.4.weight_hh_l0_reverse": { "offset": 752416, "count": 36864, "shape": [ 384, 96 ] }, "predictor.text_encoder.lstms.4.bias_ih_l0_reverse": { "offset": 789280, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.4.bias_hh_l0_reverse": { "offset": 789664, "count": 384, "shape": [ 384 ] }, "predictor.text_encoder.lstms.5.fc.weight": { "offset": 790048, "count": 12288, "shape": [ 384, 32 ] }, "predictor.text_encoder.lstms.5.fc.bias": { "offset": 802336, "count": 384, "shape": [ 384 ] }, "predictor.lstm.weight_ih_l0": { "offset": 802720, "count": 86016, "shape": [ 384, 224 ] }, "predictor.lstm.weight_hh_l0": { "offset": 888736, "count": 36864, "shape": [ 384, 96 ] }, "predictor.lstm.bias_ih_l0": { "offset": 925600, "count": 384, "shape": [ 384 ] }, "predictor.lstm.bias_hh_l0": { "offset": 925984, "count": 384, "shape": [ 384 ] }, "predictor.lstm.weight_ih_l0_reverse": { "offset": 926368, "count": 86016, "shape": [ 384, 224 ] }, "predictor.lstm.weight_hh_l0_reverse": { "offset": 1012384, "count": 36864, "shape": [ 384, 96 ] }, "predictor.lstm.bias_ih_l0_reverse": { "offset": 1049248, "count": 384, "shape": [ 384 ] }, "predictor.lstm.bias_hh_l0_reverse": { "offset": 1049632, "count": 384, "shape": [ 384 ] }, "predictor.duration_proj.linear_layer.weight": { "offset": 1050016, "count": 9600, "shape": [ 50, 192 ] }, "predictor.duration_proj.linear_layer.bias": { "offset": 1059616, "count": 50, "shape": [ 50 ] }, "predictor.shared.weight_ih_l0": { "offset": 1059666, "count": 86016, "shape": [ 384, 224 ] }, "predictor.shared.weight_hh_l0": { "offset": 1145682, "count": 36864, "shape": [ 384, 96 ] }, "predictor.shared.bias_ih_l0": { "offset": 1182546, "count": 384, "shape": [ 384 ] }, "predictor.shared.bias_hh_l0": { "offset": 1182930, "count": 384, "shape": [ 384 ] }, "predictor.shared.weight_ih_l0_reverse": { "offset": 1183314, "count": 86016, "shape": [ 384, 224 ] }, "predictor.shared.weight_hh_l0_reverse": { "offset": 1269330, "count": 36864, "shape": [ 384, 96 ] }, "predictor.shared.bias_ih_l0_reverse": { "offset": 1306194, "count": 384, "shape": [ 384 ] }, "predictor.shared.bias_hh_l0_reverse": { "offset": 1306578, "count": 384, "shape": [ 384 ] }, "text_encoder.embedding.weight": { "offset": 1306962, "count": 34176, "shape": [ 178, 192 ] }, "text_encoder.lstm.weight_ih_l0": { "offset": 1341138, "count": 73728, "shape": [ 384, 192 ] }, "text_encoder.lstm.weight_hh_l0": { "offset": 1414866, "count": 36864, "shape": [ 384, 96 ] }, "text_encoder.lstm.bias_ih_l0": { "offset": 1451730, "count": 384, "shape": [ 384 ] }, "text_encoder.lstm.bias_hh_l0": { "offset": 1452114, "count": 384, "shape": [ 384 ] }, "text_encoder.lstm.weight_ih_l0_reverse": { "offset": 1452498, "count": 73728, "shape": [ 384, 192 ] }, "text_encoder.lstm.weight_hh_l0_reverse": { "offset": 1526226, "count": 36864, "shape": [ 384, 96 ] }, "text_encoder.lstm.bias_ih_l0_reverse": { "offset": 1563090, "count": 384, "shape": [ 384 ] }, "text_encoder.lstm.bias_hh_l0_reverse": { "offset": 1563474, "count": 384, "shape": [ 384 ] }, "asr_proj.0.weight": { "offset": 1563858, "count": 98304, "shape": [ 512, 192 ] }, "asr_proj.0.bias": { "offset": 1662162, "count": 512, "shape": [ 512 ] }, "asr_proj.2.weight": { "offset": 1662674, "count": 262144, "shape": [ 512, 512 ] }, "asr_proj.2.bias": { "offset": 1924818, "count": 512, "shape": [ 512 ] }, "decoder.style": { "offset": 1925330, "count": 16, "shape": [ 1, 16 ] }, "decoder.generator.m_source.l_linear.weight": { "offset": 1925346, "count": 9, "shape": [ 1, 9 ] }, "decoder.generator.m_source.l_linear.bias": { "offset": 1925355, "count": 1, "shape": [ 1 ] } }