tpu支持

This commit is contained in:
Zchen
2025-10-12 18:41:26 +08:00
parent 1a906d3248
commit 40e4d00576
5 changed files with 456 additions and 226 deletions

File diff suppressed because one or more lines are too long

View File

@@ -0,0 +1,436 @@
2025-10-12 18:35:01,395: Using device: cpu
2025-10-12 18:35:01,395: Accelerator state: Distributed environment: NO
Num processes: 1
Process index: 0
Local process index: 0
Device: cpu
Mixed precision type: bf16
2025-10-12 18:35:38,176: torch.compile disabled for new TripleGRUDecoder compatibility
2025-10-12 18:35:38,176: Initialized RNN decoding model
2025-10-12 18:35:38,176: TripleGRUDecoder(
(noise_model): NoiseModel(
(day_layer_activation): Softsign()
(day_weights): ParameterList(
(0): Parameter containing: [torch.float32 of size 512x512]
(1): Parameter containing: [torch.float32 of size 512x512]
(2): Parameter containing: [torch.float32 of size 512x512]
(3): Parameter containing: [torch.float32 of size 512x512]
(4): Parameter containing: [torch.float32 of size 512x512]
(5): Parameter containing: [torch.float32 of size 512x512]
(6): Parameter containing: [torch.float32 of size 512x512]
(7): Parameter containing: [torch.float32 of size 512x512]
(8): Parameter containing: [torch.float32 of size 512x512]
(9): Parameter containing: [torch.float32 of size 512x512]
(10): Parameter containing: [torch.float32 of size 512x512]
(11): Parameter containing: [torch.float32 of size 512x512]
(12): Parameter containing: [torch.float32 of size 512x512]
(13): Parameter containing: [torch.float32 of size 512x512]
(14): Parameter containing: [torch.float32 of size 512x512]
(15): Parameter containing: [torch.float32 of size 512x512]
(16): Parameter containing: [torch.float32 of size 512x512]
(17): Parameter containing: [torch.float32 of size 512x512]
(18): Parameter containing: [torch.float32 of size 512x512]
(19): Parameter containing: [torch.float32 of size 512x512]
(20): Parameter containing: [torch.float32 of size 512x512]
(21): Parameter containing: [torch.float32 of size 512x512]
(22): Parameter containing: [torch.float32 of size 512x512]
(23): Parameter containing: [torch.float32 of size 512x512]
(24): Parameter containing: [torch.float32 of size 512x512]
(25): Parameter containing: [torch.float32 of size 512x512]
(26): Parameter containing: [torch.float32 of size 512x512]
(27): Parameter containing: [torch.float32 of size 512x512]
(28): Parameter containing: [torch.float32 of size 512x512]
(29): Parameter containing: [torch.float32 of size 512x512]
(30): Parameter containing: [torch.float32 of size 512x512]
(31): Parameter containing: [torch.float32 of size 512x512]
(32): Parameter containing: [torch.float32 of size 512x512]
(33): Parameter containing: [torch.float32 of size 512x512]
(34): Parameter containing: [torch.float32 of size 512x512]
(35): Parameter containing: [torch.float32 of size 512x512]
(36): Parameter containing: [torch.float32 of size 512x512]
(37): Parameter containing: [torch.float32 of size 512x512]
(38): Parameter containing: [torch.float32 of size 512x512]
(39): Parameter containing: [torch.float32 of size 512x512]
(40): Parameter containing: [torch.float32 of size 512x512]
(41): Parameter containing: [torch.float32 of size 512x512]
(42): Parameter containing: [torch.float32 of size 512x512]
(43): Parameter containing: [torch.float32 of size 512x512]
(44): Parameter containing: [torch.float32 of size 512x512]
)
(day_biases): ParameterList(
(0): Parameter containing: [torch.float32 of size 1x512]
(1): Parameter containing: [torch.float32 of size 1x512]
(2): Parameter containing: [torch.float32 of size 1x512]
(3): Parameter containing: [torch.float32 of size 1x512]
(4): Parameter containing: [torch.float32 of size 1x512]
(5): Parameter containing: [torch.float32 of size 1x512]
(6): Parameter containing: [torch.float32 of size 1x512]
(7): Parameter containing: [torch.float32 of size 1x512]
(8): Parameter containing: [torch.float32 of size 1x512]
(9): Parameter containing: [torch.float32 of size 1x512]
(10): Parameter containing: [torch.float32 of size 1x512]
(11): Parameter containing: [torch.float32 of size 1x512]
(12): Parameter containing: [torch.float32 of size 1x512]
(13): Parameter containing: [torch.float32 of size 1x512]
(14): Parameter containing: [torch.float32 of size 1x512]
(15): Parameter containing: [torch.float32 of size 1x512]
(16): Parameter containing: [torch.float32 of size 1x512]
(17): Parameter containing: [torch.float32 of size 1x512]
(18): Parameter containing: [torch.float32 of size 1x512]
(19): Parameter containing: [torch.float32 of size 1x512]
(20): Parameter containing: [torch.float32 of size 1x512]
(21): Parameter containing: [torch.float32 of size 1x512]
(22): Parameter containing: [torch.float32 of size 1x512]
(23): Parameter containing: [torch.float32 of size 1x512]
(24): Parameter containing: [torch.float32 of size 1x512]
(25): Parameter containing: [torch.float32 of size 1x512]
(26): Parameter containing: [torch.float32 of size 1x512]
(27): Parameter containing: [torch.float32 of size 1x512]
(28): Parameter containing: [torch.float32 of size 1x512]
(29): Parameter containing: [torch.float32 of size 1x512]
(30): Parameter containing: [torch.float32 of size 1x512]
(31): Parameter containing: [torch.float32 of size 1x512]
(32): Parameter containing: [torch.float32 of size 1x512]
(33): Parameter containing: [torch.float32 of size 1x512]
(34): Parameter containing: [torch.float32 of size 1x512]
(35): Parameter containing: [torch.float32 of size 1x512]
(36): Parameter containing: [torch.float32 of size 1x512]
(37): Parameter containing: [torch.float32 of size 1x512]
(38): Parameter containing: [torch.float32 of size 1x512]
(39): Parameter containing: [torch.float32 of size 1x512]
(40): Parameter containing: [torch.float32 of size 1x512]
(41): Parameter containing: [torch.float32 of size 1x512]
(42): Parameter containing: [torch.float32 of size 1x512]
(43): Parameter containing: [torch.float32 of size 1x512]
(44): Parameter containing: [torch.float32 of size 1x512]
)
(day_layer_dropout): Dropout(p=0.2, inplace=False)
(gru): GRU(7168, 7168, num_layers=2, batch_first=True, dropout=0.4)
)
(clean_speech_model): CleanSpeechModel(
(day_layer_activation): Softsign()
(day_weights): ParameterList(
(0): Parameter containing: [torch.float32 of size 512x512]
(1): Parameter containing: [torch.float32 of size 512x512]
(2): Parameter containing: [torch.float32 of size 512x512]
(3): Parameter containing: [torch.float32 of size 512x512]
(4): Parameter containing: [torch.float32 of size 512x512]
(5): Parameter containing: [torch.float32 of size 512x512]
(6): Parameter containing: [torch.float32 of size 512x512]
(7): Parameter containing: [torch.float32 of size 512x512]
(8): Parameter containing: [torch.float32 of size 512x512]
(9): Parameter containing: [torch.float32 of size 512x512]
(10): Parameter containing: [torch.float32 of size 512x512]
(11): Parameter containing: [torch.float32 of size 512x512]
(12): Parameter containing: [torch.float32 of size 512x512]
(13): Parameter containing: [torch.float32 of size 512x512]
(14): Parameter containing: [torch.float32 of size 512x512]
(15): Parameter containing: [torch.float32 of size 512x512]
(16): Parameter containing: [torch.float32 of size 512x512]
(17): Parameter containing: [torch.float32 of size 512x512]
(18): Parameter containing: [torch.float32 of size 512x512]
(19): Parameter containing: [torch.float32 of size 512x512]
(20): Parameter containing: [torch.float32 of size 512x512]
(21): Parameter containing: [torch.float32 of size 512x512]
(22): Parameter containing: [torch.float32 of size 512x512]
(23): Parameter containing: [torch.float32 of size 512x512]
(24): Parameter containing: [torch.float32 of size 512x512]
(25): Parameter containing: [torch.float32 of size 512x512]
(26): Parameter containing: [torch.float32 of size 512x512]
(27): Parameter containing: [torch.float32 of size 512x512]
(28): Parameter containing: [torch.float32 of size 512x512]
(29): Parameter containing: [torch.float32 of size 512x512]
(30): Parameter containing: [torch.float32 of size 512x512]
(31): Parameter containing: [torch.float32 of size 512x512]
(32): Parameter containing: [torch.float32 of size 512x512]
(33): Parameter containing: [torch.float32 of size 512x512]
(34): Parameter containing: [torch.float32 of size 512x512]
(35): Parameter containing: [torch.float32 of size 512x512]
(36): Parameter containing: [torch.float32 of size 512x512]
(37): Parameter containing: [torch.float32 of size 512x512]
(38): Parameter containing: [torch.float32 of size 512x512]
(39): Parameter containing: [torch.float32 of size 512x512]
(40): Parameter containing: [torch.float32 of size 512x512]
(41): Parameter containing: [torch.float32 of size 512x512]
(42): Parameter containing: [torch.float32 of size 512x512]
(43): Parameter containing: [torch.float32 of size 512x512]
(44): Parameter containing: [torch.float32 of size 512x512]
)
(day_biases): ParameterList(
(0): Parameter containing: [torch.float32 of size 1x512]
(1): Parameter containing: [torch.float32 of size 1x512]
(2): Parameter containing: [torch.float32 of size 1x512]
(3): Parameter containing: [torch.float32 of size 1x512]
(4): Parameter containing: [torch.float32 of size 1x512]
(5): Parameter containing: [torch.float32 of size 1x512]
(6): Parameter containing: [torch.float32 of size 1x512]
(7): Parameter containing: [torch.float32 of size 1x512]
(8): Parameter containing: [torch.float32 of size 1x512]
(9): Parameter containing: [torch.float32 of size 1x512]
(10): Parameter containing: [torch.float32 of size 1x512]
(11): Parameter containing: [torch.float32 of size 1x512]
(12): Parameter containing: [torch.float32 of size 1x512]
(13): Parameter containing: [torch.float32 of size 1x512]
(14): Parameter containing: [torch.float32 of size 1x512]
(15): Parameter containing: [torch.float32 of size 1x512]
(16): Parameter containing: [torch.float32 of size 1x512]
(17): Parameter containing: [torch.float32 of size 1x512]
(18): Parameter containing: [torch.float32 of size 1x512]
(19): Parameter containing: [torch.float32 of size 1x512]
(20): Parameter containing: [torch.float32 of size 1x512]
(21): Parameter containing: [torch.float32 of size 1x512]
(22): Parameter containing: [torch.float32 of size 1x512]
(23): Parameter containing: [torch.float32 of size 1x512]
(24): Parameter containing: [torch.float32 of size 1x512]
(25): Parameter containing: [torch.float32 of size 1x512]
(26): Parameter containing: [torch.float32 of size 1x512]
(27): Parameter containing: [torch.float32 of size 1x512]
(28): Parameter containing: [torch.float32 of size 1x512]
(29): Parameter containing: [torch.float32 of size 1x512]
(30): Parameter containing: [torch.float32 of size 1x512]
(31): Parameter containing: [torch.float32 of size 1x512]
(32): Parameter containing: [torch.float32 of size 1x512]
(33): Parameter containing: [torch.float32 of size 1x512]
(34): Parameter containing: [torch.float32 of size 1x512]
(35): Parameter containing: [torch.float32 of size 1x512]
(36): Parameter containing: [torch.float32 of size 1x512]
(37): Parameter containing: [torch.float32 of size 1x512]
(38): Parameter containing: [torch.float32 of size 1x512]
(39): Parameter containing: [torch.float32 of size 1x512]
(40): Parameter containing: [torch.float32 of size 1x512]
(41): Parameter containing: [torch.float32 of size 1x512]
(42): Parameter containing: [torch.float32 of size 1x512]
(43): Parameter containing: [torch.float32 of size 1x512]
(44): Parameter containing: [torch.float32 of size 1x512]
)
(day_layer_dropout): Dropout(p=0.2, inplace=False)
(gru): GRU(7168, 768, num_layers=3, batch_first=True, dropout=0.4)
(out): Linear(in_features=768, out_features=41, bias=True)
)
(noisy_speech_model): NoisySpeechModel(
(gru): GRU(7168, 768, num_layers=2, batch_first=True, dropout=0.4)
(out): Linear(in_features=768, out_features=41, bias=True)
)
)
2025-10-12 18:35:38,190: Model has 687,568,466 parameters
2025-10-12 18:35:38,191: Model has 23,639,040 day-specific parameters | 3.44% of total parameters
2025-10-12 18:38:50,217: Using device: cpu
2025-10-12 18:38:50,217: Accelerator state: Distributed environment: NO
Num processes: 1
Process index: 0
Local process index: 0
Device: cpu
Mixed precision type: bf16
2025-10-12 18:39:25,946: torch.compile disabled for new TripleGRUDecoder compatibility
2025-10-12 18:39:25,946: Initialized RNN decoding model
2025-10-12 18:39:25,946: TripleGRUDecoder(
(noise_model): NoiseModel(
(day_layer_activation): Softsign()
(day_weights): ParameterList(
(0): Parameter containing: [torch.float32 of size 512x512]
(1): Parameter containing: [torch.float32 of size 512x512]
(2): Parameter containing: [torch.float32 of size 512x512]
(3): Parameter containing: [torch.float32 of size 512x512]
(4): Parameter containing: [torch.float32 of size 512x512]
(5): Parameter containing: [torch.float32 of size 512x512]
(6): Parameter containing: [torch.float32 of size 512x512]
(7): Parameter containing: [torch.float32 of size 512x512]
(8): Parameter containing: [torch.float32 of size 512x512]
(9): Parameter containing: [torch.float32 of size 512x512]
(10): Parameter containing: [torch.float32 of size 512x512]
(11): Parameter containing: [torch.float32 of size 512x512]
(12): Parameter containing: [torch.float32 of size 512x512]
(13): Parameter containing: [torch.float32 of size 512x512]
(14): Parameter containing: [torch.float32 of size 512x512]
(15): Parameter containing: [torch.float32 of size 512x512]
(16): Parameter containing: [torch.float32 of size 512x512]
(17): Parameter containing: [torch.float32 of size 512x512]
(18): Parameter containing: [torch.float32 of size 512x512]
(19): Parameter containing: [torch.float32 of size 512x512]
(20): Parameter containing: [torch.float32 of size 512x512]
(21): Parameter containing: [torch.float32 of size 512x512]
(22): Parameter containing: [torch.float32 of size 512x512]
(23): Parameter containing: [torch.float32 of size 512x512]
(24): Parameter containing: [torch.float32 of size 512x512]
(25): Parameter containing: [torch.float32 of size 512x512]
(26): Parameter containing: [torch.float32 of size 512x512]
(27): Parameter containing: [torch.float32 of size 512x512]
(28): Parameter containing: [torch.float32 of size 512x512]
(29): Parameter containing: [torch.float32 of size 512x512]
(30): Parameter containing: [torch.float32 of size 512x512]
(31): Parameter containing: [torch.float32 of size 512x512]
(32): Parameter containing: [torch.float32 of size 512x512]
(33): Parameter containing: [torch.float32 of size 512x512]
(34): Parameter containing: [torch.float32 of size 512x512]
(35): Parameter containing: [torch.float32 of size 512x512]
(36): Parameter containing: [torch.float32 of size 512x512]
(37): Parameter containing: [torch.float32 of size 512x512]
(38): Parameter containing: [torch.float32 of size 512x512]
(39): Parameter containing: [torch.float32 of size 512x512]
(40): Parameter containing: [torch.float32 of size 512x512]
(41): Parameter containing: [torch.float32 of size 512x512]
(42): Parameter containing: [torch.float32 of size 512x512]
(43): Parameter containing: [torch.float32 of size 512x512]
(44): Parameter containing: [torch.float32 of size 512x512]
)
(day_biases): ParameterList(
(0): Parameter containing: [torch.float32 of size 1x512]
(1): Parameter containing: [torch.float32 of size 1x512]
(2): Parameter containing: [torch.float32 of size 1x512]
(3): Parameter containing: [torch.float32 of size 1x512]
(4): Parameter containing: [torch.float32 of size 1x512]
(5): Parameter containing: [torch.float32 of size 1x512]
(6): Parameter containing: [torch.float32 of size 1x512]
(7): Parameter containing: [torch.float32 of size 1x512]
(8): Parameter containing: [torch.float32 of size 1x512]
(9): Parameter containing: [torch.float32 of size 1x512]
(10): Parameter containing: [torch.float32 of size 1x512]
(11): Parameter containing: [torch.float32 of size 1x512]
(12): Parameter containing: [torch.float32 of size 1x512]
(13): Parameter containing: [torch.float32 of size 1x512]
(14): Parameter containing: [torch.float32 of size 1x512]
(15): Parameter containing: [torch.float32 of size 1x512]
(16): Parameter containing: [torch.float32 of size 1x512]
(17): Parameter containing: [torch.float32 of size 1x512]
(18): Parameter containing: [torch.float32 of size 1x512]
(19): Parameter containing: [torch.float32 of size 1x512]
(20): Parameter containing: [torch.float32 of size 1x512]
(21): Parameter containing: [torch.float32 of size 1x512]
(22): Parameter containing: [torch.float32 of size 1x512]
(23): Parameter containing: [torch.float32 of size 1x512]
(24): Parameter containing: [torch.float32 of size 1x512]
(25): Parameter containing: [torch.float32 of size 1x512]
(26): Parameter containing: [torch.float32 of size 1x512]
(27): Parameter containing: [torch.float32 of size 1x512]
(28): Parameter containing: [torch.float32 of size 1x512]
(29): Parameter containing: [torch.float32 of size 1x512]
(30): Parameter containing: [torch.float32 of size 1x512]
(31): Parameter containing: [torch.float32 of size 1x512]
(32): Parameter containing: [torch.float32 of size 1x512]
(33): Parameter containing: [torch.float32 of size 1x512]
(34): Parameter containing: [torch.float32 of size 1x512]
(35): Parameter containing: [torch.float32 of size 1x512]
(36): Parameter containing: [torch.float32 of size 1x512]
(37): Parameter containing: [torch.float32 of size 1x512]
(38): Parameter containing: [torch.float32 of size 1x512]
(39): Parameter containing: [torch.float32 of size 1x512]
(40): Parameter containing: [torch.float32 of size 1x512]
(41): Parameter containing: [torch.float32 of size 1x512]
(42): Parameter containing: [torch.float32 of size 1x512]
(43): Parameter containing: [torch.float32 of size 1x512]
(44): Parameter containing: [torch.float32 of size 1x512]
)
(day_layer_dropout): Dropout(p=0.2, inplace=False)
(gru): GRU(7168, 7168, num_layers=2, batch_first=True, dropout=0.4)
)
(clean_speech_model): CleanSpeechModel(
(day_layer_activation): Softsign()
(day_weights): ParameterList(
(0): Parameter containing: [torch.float32 of size 512x512]
(1): Parameter containing: [torch.float32 of size 512x512]
(2): Parameter containing: [torch.float32 of size 512x512]
(3): Parameter containing: [torch.float32 of size 512x512]
(4): Parameter containing: [torch.float32 of size 512x512]
(5): Parameter containing: [torch.float32 of size 512x512]
(6): Parameter containing: [torch.float32 of size 512x512]
(7): Parameter containing: [torch.float32 of size 512x512]
(8): Parameter containing: [torch.float32 of size 512x512]
(9): Parameter containing: [torch.float32 of size 512x512]
(10): Parameter containing: [torch.float32 of size 512x512]
(11): Parameter containing: [torch.float32 of size 512x512]
(12): Parameter containing: [torch.float32 of size 512x512]
(13): Parameter containing: [torch.float32 of size 512x512]
(14): Parameter containing: [torch.float32 of size 512x512]
(15): Parameter containing: [torch.float32 of size 512x512]
(16): Parameter containing: [torch.float32 of size 512x512]
(17): Parameter containing: [torch.float32 of size 512x512]
(18): Parameter containing: [torch.float32 of size 512x512]
(19): Parameter containing: [torch.float32 of size 512x512]
(20): Parameter containing: [torch.float32 of size 512x512]
(21): Parameter containing: [torch.float32 of size 512x512]
(22): Parameter containing: [torch.float32 of size 512x512]
(23): Parameter containing: [torch.float32 of size 512x512]
(24): Parameter containing: [torch.float32 of size 512x512]
(25): Parameter containing: [torch.float32 of size 512x512]
(26): Parameter containing: [torch.float32 of size 512x512]
(27): Parameter containing: [torch.float32 of size 512x512]
(28): Parameter containing: [torch.float32 of size 512x512]
(29): Parameter containing: [torch.float32 of size 512x512]
(30): Parameter containing: [torch.float32 of size 512x512]
(31): Parameter containing: [torch.float32 of size 512x512]
(32): Parameter containing: [torch.float32 of size 512x512]
(33): Parameter containing: [torch.float32 of size 512x512]
(34): Parameter containing: [torch.float32 of size 512x512]
(35): Parameter containing: [torch.float32 of size 512x512]
(36): Parameter containing: [torch.float32 of size 512x512]
(37): Parameter containing: [torch.float32 of size 512x512]
(38): Parameter containing: [torch.float32 of size 512x512]
(39): Parameter containing: [torch.float32 of size 512x512]
(40): Parameter containing: [torch.float32 of size 512x512]
(41): Parameter containing: [torch.float32 of size 512x512]
(42): Parameter containing: [torch.float32 of size 512x512]
(43): Parameter containing: [torch.float32 of size 512x512]
(44): Parameter containing: [torch.float32 of size 512x512]
)
(day_biases): ParameterList(
(0): Parameter containing: [torch.float32 of size 1x512]
(1): Parameter containing: [torch.float32 of size 1x512]
(2): Parameter containing: [torch.float32 of size 1x512]
(3): Parameter containing: [torch.float32 of size 1x512]
(4): Parameter containing: [torch.float32 of size 1x512]
(5): Parameter containing: [torch.float32 of size 1x512]
(6): Parameter containing: [torch.float32 of size 1x512]
(7): Parameter containing: [torch.float32 of size 1x512]
(8): Parameter containing: [torch.float32 of size 1x512]
(9): Parameter containing: [torch.float32 of size 1x512]
(10): Parameter containing: [torch.float32 of size 1x512]
(11): Parameter containing: [torch.float32 of size 1x512]
(12): Parameter containing: [torch.float32 of size 1x512]
(13): Parameter containing: [torch.float32 of size 1x512]
(14): Parameter containing: [torch.float32 of size 1x512]
(15): Parameter containing: [torch.float32 of size 1x512]
(16): Parameter containing: [torch.float32 of size 1x512]
(17): Parameter containing: [torch.float32 of size 1x512]
(18): Parameter containing: [torch.float32 of size 1x512]
(19): Parameter containing: [torch.float32 of size 1x512]
(20): Parameter containing: [torch.float32 of size 1x512]
(21): Parameter containing: [torch.float32 of size 1x512]
(22): Parameter containing: [torch.float32 of size 1x512]
(23): Parameter containing: [torch.float32 of size 1x512]
(24): Parameter containing: [torch.float32 of size 1x512]
(25): Parameter containing: [torch.float32 of size 1x512]
(26): Parameter containing: [torch.float32 of size 1x512]
(27): Parameter containing: [torch.float32 of size 1x512]
(28): Parameter containing: [torch.float32 of size 1x512]
(29): Parameter containing: [torch.float32 of size 1x512]
(30): Parameter containing: [torch.float32 of size 1x512]
(31): Parameter containing: [torch.float32 of size 1x512]
(32): Parameter containing: [torch.float32 of size 1x512]
(33): Parameter containing: [torch.float32 of size 1x512]
(34): Parameter containing: [torch.float32 of size 1x512]
(35): Parameter containing: [torch.float32 of size 1x512]
(36): Parameter containing: [torch.float32 of size 1x512]
(37): Parameter containing: [torch.float32 of size 1x512]
(38): Parameter containing: [torch.float32 of size 1x512]
(39): Parameter containing: [torch.float32 of size 1x512]
(40): Parameter containing: [torch.float32 of size 1x512]
(41): Parameter containing: [torch.float32 of size 1x512]
(42): Parameter containing: [torch.float32 of size 1x512]
(43): Parameter containing: [torch.float32 of size 1x512]
(44): Parameter containing: [torch.float32 of size 1x512]
)
(day_layer_dropout): Dropout(p=0.2, inplace=False)
(gru): GRU(7168, 768, num_layers=3, batch_first=True, dropout=0.4)
(out): Linear(in_features=768, out_features=41, bias=True)
)
(noisy_speech_model): NoisySpeechModel(
(gru): GRU(7168, 768, num_layers=2, batch_first=True, dropout=0.4)
(out): Linear(in_features=768, out_features=41, bias=True)
)
)
2025-10-12 18:39:25,958: Model has 687,568,466 parameters
2025-10-12 18:39:25,958: Model has 23,639,040 day-specific parameters | 3.44% of total parameters