Using Landsat 5 images with v0.1 to get embeddings.

[ritwikvashistha]

Hi,

I am trying to use v0.1 to get embeddings for Landsat 5 images. I know it's not part of the data module, but I have some geotiffs on my machine that I was hoping to pass to the model and get the embeddings.
However, I ran into some issues and wasn't able to get the embeddings. I am using the tutorial on embeddings as a reference.

The error is get is:

KeyError                                  Traceback (most recent call last)
Cell In[30], line 3
      1 embeddings = []
----> 3 for batch in trn_dl:
      4     with torch.inference_mode():
      5         # Move data from to the device of model
      6         batch["pixels"] = batch["pixels"].to(rgb_model.device)

File ~/micromamba/envs/claymodel/lib/python3.11/site-packages/torch/utils/data/dataloader.py:630, in _BaseDataLoaderIter.__next__(self)
    627 if self._sampler_iter is None:
    628     # TODO(https://github.com/pytorch/pytorch/issues/76750)
    629     self._reset()  # type: ignore[call-arg]
--> 630 data = self._next_data()
    631 self._num_yielded += 1
    632 if self._dataset_kind == _DatasetKind.Iterable and \
    633         self._IterableDataset_len_called is not None and \
    634         self._num_yielded > self._IterableDataset_len_called:

File ~/micromamba/envs/claymodel/lib/python3.11/site-packages/torch/utils/data/dataloader.py:1345, in _MultiProcessingDataLoaderIter._next_data(self)
   1343 else:
   1344     del self._task_info[idx]
-> 1345     return self._process_data(data)

File ~/micromamba/envs/claymodel/lib/python3.11/site-packages/torch/utils/data/dataloader.py:1371, in _MultiProcessingDataLoaderIter._process_data(self, data)
   1369 self._try_put_index()
   1370 if isinstance(data, ExceptionWrapper):
-> 1371     data.reraise()
   1372 return data

File ~/micromamba/envs/claymodel/lib/python3.11/site-packages/torch/_utils.py:694, in ExceptionWrapper.reraise(self)
    690 except TypeError:
    691     # If the exception takes multiple arguments, don't try to
    692     # instantiate since we don't know how to
    693     raise RuntimeError(msg) from None
--> 694 raise exception

KeyError: Caught KeyError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/Users/ritwikvashistha/micromamba/envs/claymodel/lib/python3.11/site-packages/torch/utils/data/_utils/worker.py", line 308, in _worker_loop
    data = fetcher.fetch(index)
           ^^^^^^^^^^^^^^^^^^^^
  File "/Users/ritwikvashistha/micromamba/envs/claymodel/lib/python3.11/site-packages/torch/utils/data/_utils/fetch.py", line 51, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/ritwikvashistha/micromamba/envs/claymodel/lib/python3.11/site-packages/torch/utils/data/_utils/fetch.py", line 51, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
            ~~~~~~~~~~~~^^^^^
  File "/Users/ritwikvashistha/model/src/datamodule.py", line 84, in __getitem__
    cube = self.read_chip(chip_path)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/ritwikvashistha/model/src/datamodule.py", line 62, in read_chip
    date = chip.tags()["date"]  # YYYY-MM-DD
           ~~~~~~~~~~~^^^^^^^^
KeyError: 'date'

The file given to a model is indeed a geotiff downloaded from Google Earth engine and should have information such as date within the file. Here's a link to the file that I was trying to give to the model: https://drive.google.com/file/d/168LOO3cINDDoUWKNJTwTgKHRnp5PcmKj/view?usp=sharing. I would appreciate any insights about how to go about using Landsat 5 images with the model!

[weiji14]

Hi @ritwikvashistha, good to see someone trying things out on Landsat! The error you're getting is due to the DataModule expecting a date tag in the GeoTIFF file, and this tag is something we've added to our datacubes for training, since the Clay model expects a timestamp encoding (see also #70 for context). We've been debating on whether to use ACQUISITIONDATETIME or TIFFTAG_DATETIME instead actually, can you check if the Landsat TIFF you have has one of those tags, and maybe we can finally decide which one to pick 🙂