mmcdermott/EventStreamGPT
Dataset and modelling infrastructure for modelling "event streams": sequences of continuous time, multivariate events with complex internal dependencies.
Jupyter NotebookMIT
Issues
- 4
Setting min_seq_len=1 in PytorchDatasetConfig with task dataframes leads to ragged tensors index error
#118 opened by juancq - 11
- 0
- 5
Setting min_seq_len in PytorchDatasetConfig to < 2 with task dataframes leads to keyerror
#109 opened by juancq - 4
- 2
Could not override 'config.task_specific_params.pooling_method' when lauching wandb agent
#108 opened by Rhett-Ying - 1
Processing Synthetic Data with ESGPT
#113 opened by sujaybanerjee - 1
- 1
- 0
Task `to_int_index` does not pass tests with more recent versions of polars
#112 opened by mmcdermott - 2
- 1
- 1
Getting started steps
#96 opened by rvandewater - 0
- 0
PytorchDataset doesn't have a way to extract subject_ids, start_time, and end_time all together
#75 opened by pargaw - 0
Should be possible to resume pre-training and evaluate a pre-trained model post-hoc
#91 opened by mmcdermott - 2
May be able to make temporal loss more stable by directly computing log-prob from pre-activation function outputs
#93 opened by mmcdermott - 0
CATEGORICAL_ONLY v. NUMERICAL_ONLY v. CATEGORICAL_AND_NUMERICAL is confusing.
#13 opened by mmcdermott - 1
- 0
- 0
- 1
Default model should be simplified and use the built-in PyTorch transformer layers (though still support structured attention)
#10 opened by mmcdermott - 0
Torch.compile doesn't work for some reason.
#11 opened by mmcdermott - 0
- 0
- 0
- 0
- 2
- 0
- 0
Should allow use of NLL-Beta
#26 opened by mmcdermott - 6
conda environment file fails
#32 opened by juancq - 1
- 4
- 6
- 0
Modeling Configs Should be Re-arranged for Clarity, Better Defaults, and Ease-of-use
#61 opened by mmcdermott - 0
Metrics in weights and biases should use `/` not `_` to separate name parts.
#60 opened by mmcdermott - 1
Setting maximum number of threads Polars will use
#47 opened by juancq - 1
- 4
Add streaming flag for calls to polars collect
#46 opened by juancq - 0
Option of using pyarrow when writing parquet
#49 opened by juancq - 1
- 1
- 1
- 2
- 1
self has no task_df member.
#34 opened by mmcdermott - 0
- 0
Vocabulary serialization into configs should not include observation frequencies.
#42 opened by mmcdermott - 3
- 2
- 0