Document config #1600
Document config #1600
Conversation
|
Super, thanks! I will make a pass through it in the next days. |
CONFIG_DOCUMENTATION.md
Outdated
| | `start_date` | datetime | `1979-01-01T00:00` | Training data start date | | ||
| | `end_date` | datetime | `2022-12-31T00:00` | Training data end date | | ||
| | `time_window_step` | duration | `06:00:00` | Time step between windows | | ||
| | `time_window_len` | duration | `06:00:00` | Length of each time window | |
There was a problem hiding this comment.
I'd suggest changing type from "duration" to "timedelta". I think it's better to use name already known from pandas or numpy.
There was a problem hiding this comment.
feel free to edit it! I am open
There was a problem hiding this comment.
I'd love to but I don't have write access to this repo. Can I open PR into your branch?
There was a problem hiding this comment.
Yes, please open a PR against this one.
CONFIG_DOCUMENTATION.md
Outdated
| | `fe_layer_norm_after_blocks` | list | `[]` | Block indices after which to add LayerNorm (0-indexed) | | ||
| | `fe_impute_latent_noise_std` | float | `0.0` | Standard deviation for latent noise imputation | | ||
| | `forecast_att_dense_rate` | float | `1.0` | Precentage of attention layers that are dense | | ||
| | `with_step_conditioning` | bool | `True` | Enable step conditioning in forecast engine | |
There was a problem hiding this comment.
with_step_conditioning can be removed. I will submit a commit change
|
|
||
| | Parameter | Type | Default | Options | Description | | ||
| |-----------|------|---------|---------|-------------| | ||
| | `time_step` | timedelta | `06:00:00` | | Time step between forecast steps | |
There was a problem hiding this comment.
Hi, I got a bit confused about having "time step between", would it capture it better if we write:
"(Fixed) Time interval between consecutive forecast steps"?
There was a problem hiding this comment.
Happy to commit so if you agree.
Description
Generate with an AI coding agent, verified by human
Issue Number
Closes #1599
Is this PR a draft? Mark it as draft.