Quickstart¶
Run your first neuralbench evaluation in a few commands.
Which entry point?¶
The same experiments are reachable three ways – same YAML configs, same cache, same results folder. What decides between them is where your model lives and what you want handed back.
|
|
|
|
|---|---|---|---|
Names the model |
registered name ( |
registered name ( |
an |
Covers |
tasks x datasets x models x grid |
the same selections |
one instance, over tasks and datasets |
Hands back |
results on disk; |
the same, from Python |
the results, as a |
Runs on |
Slurm, or locally with |
the same |
this process, or Slurm with |
Reach for the CLI to run or reproduce the benchmark, for
run_benchmark() to drive those same runs from a script, and
for evaluate_model() while developing a model that has no
config in this repo. The three tutorials below follow that order.
Note
đź“Ą You can download these tutorials with the buttons at the bottom of each page.
