Add tutorial how to override arch params #32
Conversation
PicoCentauri
force-pushed
the
arch_override_docs
branch
2 times, most recently
from
81d2ce8
to
fdb5de8
Compare
docs/static/options.yaml
Outdated
| @@ -1,5 +1,6 @@ | |||
| defaults: | |||
| - architecture: soap_bpnn # architecture used to train the model | |||
| - _self_ | |||
|
|
|||
| # Section defining the parameters for structure and target data | |||
| dataset: | |||
There was a problem hiding this comment.
this looks quite a bit different from the previous discussion, is this the old version of the parameters?
There was a problem hiding this comment.
Yes it still builds on the current main branch, where the new option was not in until 5 minutes ago.
I will rebase and update the config.
| syntax are available at https://hydra.cc/docs/advanced/override_grammar/basic/. | ||
|
|
||
| .. note:: | ||
| `metatensor-models` always writes the fully expanded ``options.yaml`` for your |
There was a problem hiding this comment.
we should tell where the fully expanded file can be found
There was a problem hiding this comment.
Yes, good point.
There was a problem hiding this comment.
In general, I don't think we should mention the option of having the defaults override our custom config. Let's make it simple for the user: just put _self_ at the very end. I like the idea of having links to hydra for more advanced settings, but I just wouldn't mention the alternatives in our docs because they don't apply to our users
| :ref:`available-architectures` section. For SOAP-BPNN models, the parameters are | ||
| detailed at :ref:`architecture-soap-bpnn`. |
There was a problem hiding this comment.
| :ref:`available-architectures` section. For SOAP-BPNN models, the parameters are | |
| detailed at :ref:`architecture-soap-bpnn`. | |
| :ref:`available-architectures` section. For example, the parameters of the SOAP-BPNN models are | |
| detailed at :ref:`architecture-soap-bpnn`. |
| training: | ||
| num_epochs: 200 | ||
|
|
||
| # Dataset config omitted for simplicity. |
There was a problem hiding this comment.
Here we should have a very minimal dataset.
training_set: your_dataset.xyz
|
|
||
| .. note:: | ||
| `metatensor-models` always writes the fully expanded ``options.yaml`` for your | ||
| reference. |
There was a problem hiding this comment.
| reference. | |
| reference and reproducibility purposes. |
PicoCentauri
force-pushed
the
arch_override_docs
branch
from
fdb5de8
to
3d1dd91
Compare
PicoCentauri
force-pushed
the
arch_override_docs
branch
from
3d1dd91
to
9061f1c
Compare
|
I pushed an update to the tutorial. Let me know, if this is better. |
Fixes #23
📚 Documentation preview 📚: https://metatensor-models--32.org.readthedocs.build/en/32/