issues
search
eth-easl
/
modyn
Modyn is a research-platform for training ML models on growing datasets.
MIT License
22
stars
3
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Specify cuda device for model storage
#477
MaxiBoether
opened
1 month ago
0
feat: Evaluator cleanup endpoint
#476
robinholzi
closed
1 month ago
2
fix: Adjust data generation scripts
#475
robinholzi
closed
1 month ago
2
Add description and scripts for downloading and generating the CGLM dataset
#474
MaxiBoether
closed
1 month ago
3
fix: Fix csv quoting that broke some datasets
#473
robinholzi
closed
1 month ago
2
feat: Allow to pass number of expected batches or epochs to learning rate scheduler
#472
MaxiBoether
closed
1 month ago
4
Fix selection strategy on RHO pipeline
#471
XianzheMa
closed
1 month ago
2
[WIP] Implement `__len__` on OnlineDataset to allow for correct learning rate scheduling
#470
MaxiBoether
closed
1 month ago
1
feat: Support top-N Accuracy
#469
MaxiBoether
closed
1 month ago
2
refactor: Reorganize and document `pytorch_trainer`
#468
robinholzi
closed
1 month ago
3
fix: Allow using multiple F1 scores in evaluation
#467
MaxiBoether
closed
1 month ago
5
Support for RS2 at the Selector instead of downsampling
#466
MaxiBoether
opened
1 month ago
0
feat: Support for RS2 Downsampler
#465
MaxiBoether
closed
1 month ago
5
feat: Pipeline run analytics visualisation tool
#464
robinholzi
closed
1 month ago
3
feat: Adjust yearbook multi-experiment runner
#463
robinholzi
closed
1 month ago
2
feat: Enable `torch.inference_mode` if downsampler supports it
#462
MaxiBoether
closed
1 month ago
5
Move `SelectionStrategy` to root level of pipeline configuration
#461
XianzheMa
closed
1 month ago
2
Allow users to specify whether to only shuffle partition order or data within partitions
#460
MaxiBoether
opened
1 month ago
0
RHO-Loss Step 3: Train IL Model
#459
XianzheMa
closed
1 month ago
2
RHO-LOSS Step 3: train il model
#458
XianzheMa
closed
1 month ago
2
RHO-LOSS Step 1: fetch or create rho pipeline id
#457
XianzheMa
closed
1 month ago
2
feat: Shuffle between epochs
#456
MaxiBoether
closed
1 month ago
7
RHO-LOSS step 1: fetch rho pipeline id
#455
XianzheMa
closed
1 month ago
3
Use explicit pydantic model in selector business logic
#454
XianzheMa
closed
1 month ago
2
RHO-LOSS step 1: fetch or create the rho pipeline id
#453
XianzheMa
closed
1 month ago
3
feat: Add function fieldvalidators
#452
robinholzi
closed
1 month ago
2
feat: Add more powerful evaluation methods
#451
robinholzi
closed
1 month ago
6
Accumulate batches in Batch-Then-Sample mode
#450
MaxiBoether
closed
1 month ago
8
RHO-LOSS step 2: prepare the holdout set
#449
XianzheMa
closed
1 month ago
2
feat: Support running multiple experiments sequentially using a dynamic config
#448
robinholzi
closed
1 month ago
2
refactor: Use more explicit trigger configurations
#447
robinholzi
closed
1 month ago
2
fix: Fix linting, integrate `ruff` and typo check `pre-commit` checks
#446
robinholzi
closed
1 month ago
3
Change parameter list of the downsampling constructor and outlines the implementation of RHO-LOSS
#445
XianzheMa
closed
1 month ago
2
Fix downsampler update in downsampler scheduler
#444
XianzheMa
closed
1 month ago
2
Add missing `model.eval()` during BtS downsampling
#443
XianzheMa
closed
1 month ago
5
RFC: How to evaluate pipelines and models
#442
robinholzi
opened
1 month ago
4
test: Add unit tests enforcing all configs to be up to date
#441
robinholzi
closed
1 month ago
5
Log the pydantic validation error and add tokenizer to evaluation dataset schema
#440
XianzheMa
closed
2 months ago
2
feat: Adjust logging with more meta information for plotting
#439
robinholzi
closed
1 month ago
2
Fix selection strategy config
#438
XianzheMa
closed
2 months ago
3
refactor: Adjust pipeline orchestration in `PipelineExecutor`
#437
robinholzi
closed
1 month ago
9
Reject unknown properties in the pipeline config
#436
XianzheMa
closed
2 months ago
2
feat: Integrate pydantic validation into supervisor
#435
robinholzi
closed
2 months ago
3
refactor: Adjust pipeline orchestration in `PipelineExecutor`
#434
robinholzi
closed
2 months ago
1
Implement a ModelDownloader as common util
#433
jenny011
closed
2 months ago
0
fix: Make integration tests runnable locally
#432
robinholzi
closed
2 months ago
2
refactor: Use `typer` for cli
#431
robinholzi
closed
2 months ago
2
feat: Integrate pydantic validation into supervisor
#430
robinholzi
closed
2 months ago
5
Record the time span of a trigger
#429
XianzheMa
closed
2 months ago
2
Make `start_timestamp` and `end_timestamp` in evaluation gRPC requests optional
#428
XianzheMa
closed
2 months ago
2
Previous
Next