first commit

models/Chronos2.py

@@ -0,0 +1,43 @@
+import torch
+from torch import nn
+from layers.Transformer_EncDec import Encoder, EncoderLayer
+from layers.SelfAttention_Family import FullAttention, AttentionLayer
+from layers.Embed import PatchEmbedding
+from chronos import BaseChronosPipeline
+
+
+class Model(nn.Module):
+    def __init__(self, configs):
+        """
+        patch_len: int, patch len for patch_embedding
+        stride: int, stride for patch_embedding
+        """
+        super().__init__()
+        self.model = BaseChronosPipeline.from_pretrained("amazon/chronos-2", device_map="cuda")
+        self.task_name = configs.task_name
+        self.seq_len = configs.seq_len
+        self.pred_len = configs.pred_len
+
+    def forecast(self, x_enc, x_mark_enc, x_dec, x_mark_dec):
+        means = x_enc.mean(1, keepdim=True).detach()
+        x_enc = x_enc.sub(means)
+        stdev = torch.sqrt(
+            torch.var(x_enc, dim=1, keepdim=True, unbiased=False) + 1e-5)
+        x_enc = x_enc.div(stdev)
+
+        B, L, C = x_enc.shape
+        x_enc = x_enc.permute(0, 2, 1)
+        quantiles, dec_out = self.model.predict_quantiles(x_enc.cpu().numpy(), prediction_length=self.pred_len, quantile_levels=[0.1, 0.5, 0.9])
+        dec_out = torch.stack(dec_out, dim=0).to(x_enc.device)
+        dec_out= dec_out.permute(0, 2, 1)
+        dec_out = dec_out * \
+                  (stdev[:, 0, :].unsqueeze(1).repeat(1, self.pred_len, 1))
+        dec_out = dec_out + \
+                  (means[:, 0, :].unsqueeze(1).repeat(1, self.pred_len, 1))
+        return dec_out
+
+    def forward(self, x_enc, x_mark_enc, x_dec, x_mark_dec, mask=None):
+        if self.task_name == 'zero_shot_forecast':
+            dec_out = self.forecast(x_enc, x_mark_enc, x_dec, x_mark_dec)
+            return dec_out
+        return None