Dirk Groeneveld
Dirk Groeneveld
Test cases: - [ ] Direct step dependency - [ ] Step dependency through a list/dict/set/tuple - [ ] Step dependency through a complex (`FromParams`) object - [ ] Anonymous...
I have a bunch of text pairs that I want to tokenize. Some of those texts are too long for the transformer, so I ask the tokenizer to truncate with...
I'm looking at these docs: https://huggingface.co/docs/transformers/main/en/main_classes/tokenizer#transformers.PreTrainedTokenizer.encode They say you can pass in token ids instead of a string. But when you try you get a `TypeError`: ``` In [2]: import...
### 🚀 The feature, motivation and pitch Beaker experiments have significant overhead. When we're running many small Tango steps, we can save some time by running multiple steps at once....
### 🐛 Describe the bug Run the catwalk training job specified here: https://github.com/allenai/catwalk/commit/5ba019204b0ff36c1c4da7feab4515342e9d9ad2 Command line is `tango --settings experiments/train_all_the_things/tango.yml run experiments/train_all_the_things/train_all_the_things.jsonnet`. It will run for quite a while. Two jobs...
We write out optimizer state for step0, but it doesn't make sense to have it. Thankfully the file we write is small, but it actually crashes when you try to...
It is suspicious that we had two slightly different models (one with biases, one without), that both spiked at exactly the same moment. This suggests there might be a data...
What happens now === Our runs produce "checkpoint directories". You might have seen them. Checkpoint directories contain a bunch of debris from a run, including between 0 and n actual...