init main with tspulse-hybrid-allhead-512-p8-r1

config.json

@@ -0,0 +1,175 @@
+{
+  "architectures": [
+    "TSPulseForReconstruction"
+  ],
+  "batch_aware_masking": false,
+  "categorical_vocab_size_list": null,
+  "channel_consistent_masking": true,
+  "channel_mix_init": "identity",
+  "channel_register_tokens": null,
+  "channel_virtual_expand_scale": 1,
+  "classification_mode": "full_embedding",
+  "context_length": 512,
+  "d_model": 24,
+  "d_model_layerwise": [
+    24,
+    24,
+    24,
+    24,
+    24,
+    24,
+    24,
+    24
+  ],
+  "d_model_layerwise_scale": [
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1
+  ],
+  "d_model_scale": 3,
+  "data_actual_context_length": null,
+  "decoder_d_model": 24,
+  "decoder_d_model_layerwise": [
+    24,
+    24
+  ],
+  "decoder_d_model_layerwise_scale": [
+    1,
+    1
+  ],
+  "decoder_d_model_scale": 3,
+  "decoder_mode": "common_channel",
+  "decoder_num_channels_layerwise": [
+    1,
+    1
+  ],
+  "decoder_num_channels_layerwise_scale": [
+    1,
+    1
+  ],
+  "decoder_num_layers": 2,
+  "decoder_num_patches_layerwise": [
+    128,
+    128
+  ],
+  "decoder_num_patches_layerwise_scale": [
+    1,
+    1
+  ],
+  "disable_mask_in_classification_eval": false,
+  "dropout": 0.2,
+  "enable_fft_prob_loss": true,
+  "expansion_factor": 2,
+  "fft_applied_on": "scaled_ts",
+  "fft_mask_ratio": null,
+  "fft_mask_strategy": "magnitude",
+  "fft_original_signal_loss_weight": 1,
+  "fft_prob_length": null,
+  "fft_prob_mode": "log",
+  "fft_prob_weight": 1,
+  "fft_remove_component": "last",
+  "fft_time_add_forecasting_pt_loss": true,
+  "fft_time_add_forecasting_pt_loss_weight": 1,
+  "fft_time_consistent_masking": true,
+  "fft_weight": 1,
+  "free_channel_flow": true,
+  "full_patch_mask_percentage": 0.5,
+  "fuse_fft": true,
+  "gated_attention_activation": "softmax",
+  "gated_attn": true,
+  "head_aggregation": null,
+  "head_aggregation_dim": "patch",
+  "head_attention": false,
+  "head_dropout": 0.2,
+  "head_gated_attention_activation": "softmax",
+  "head_reduce_channels": null,
+  "head_reduce_d_model": null,
+  "hydra_class_attention": true,
+  "hydra_class_head": null,
+  "init_embed": "pytorch",
+  "init_linear": "pytorch",
+  "init_processing": true,
+  "init_std": 0.02,
+  "loss": "mse",
+  "loss_apply_mode": "mask",
+  "mask_block_length": 8,
+  "mask_ratio": 0.7,
+  "mask_type": "var_hybrid",
+  "masked_reconstruction_loss_weight": 1,
+  "min_allowed_range": 4,
+  "minimum_scale": 0.001,
+  "mode": "common_channel",
+  "model_type": "tspulse",
+  "norm_eps": 1e-05,
+  "norm_mlp": "LayerNorm",
+  "num_channels_layerwise": [
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1
+  ],
+  "num_channels_layerwise_scale": [
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1
+  ],
+  "num_full_patches_for_hybrid_mask": 2,
+  "num_input_channels": 1,
+  "num_layers": 8,
+  "num_patches": 128,
+  "num_patches_layerwise": [
+    128,
+    128,
+    128,
+    128,
+    128,
+    128,
+    128,
+    128
+  ],
+  "num_patches_layerwise_scale": [
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1,
+    1
+  ],
+  "num_targets": 3,
+  "output_range": null,
+  "patch_last": true,
+  "patch_length": 8,
+  "patch_register_tokens": 10,
+  "patch_stride": 8,
+  "positional_encoding_type": "sincos",
+  "post_init": false,
+  "prediction_length": 16,
+  "reconstruction_loss_weight": 1,
+  "reconstruction_type": "patchwise",
+  "register_mixer_layers": 3,
+  "revin_affine": true,
+  "scaling": "revin",
+  "self_attn": false,
+  "self_attn_heads": 1,
+  "stride_ratio": 1,
+  "torch_dtype": "float32",
+  "transformers_version": "4.38.0",
+  "use_learnable_mask_token": true,
+  "use_positional_encoding": false
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57fa03b67d1473a7253ac37801b24c391d1fa931395db986743e2f59e556b9ac
+size 4374640