40 lines
1.4 KiB
Python
40 lines
1.4 KiB
Python
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-bolt-base",
|
|
device_map="cuda", # use "cpu" for CPU inference and "mps" for Apple Silicon
|
|
torch_dtype=torch.bfloat16,
|
|
)
|
|
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):
|
|
outputs = []
|
|
for i in range(x_enc.shape[-1]):
|
|
output = self.model.predict(x_enc[...,i], prediction_length=self.pred_len)
|
|
output = output.mean(dim=1)
|
|
outputs.append(output)
|
|
dec_out = torch.stack(outputs, dim=-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
|