composer
composer copied to clipboard
mosaicml
Supercharge Your Model Training
** Environment ** - OS: Ubuntu 22.04 LTS - GPU: ``` *-display description: VGA compatible controller product: GP102 [GeForce GTX 1080 Ti] vendor: NVIDIA Corporation physical id: 0 bus info:...
I have a few questions about the EMA docs: https://docs.mosaicml.com/en/latest/method_cards/ema.html - Will the EMA version of the model automatically be used during inference time? - What if I want to...
Uses evaluation logic pseudocode to provide a better intuition for the logic and the relationship between `ComposerModel.validate` and `ComposerModel.metrics`. Adds some exposition around using the `Evaluators` class. Closes CO-861
I'm trying to do something simple: log validation loss every N batches. Reading through the docs I find this: ``` def validate(self, batch): labels = batch.pop('labels') output = self.forward(batch) output...
- Runs fine tuning in 2 rounds of parallel jobs (as opposed to `n_pretrain_checkpoints * 2` rounds of parallel jobs). This enables more efficient parallelization, especially when running on machines...
Making this a draft PR in case I'm biting off more than I can chew... ### This PR does the folllowing: * Simplify typing of `console_log_level: Union[str, int, LogLevel]` because...
Closes [JIRAs](https://mosaicml.atlassian.net/browse/CO-680). Putting this up as a reference during refactor discussions
