SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

feat(wandb): logging and configuration improvements (#10826)

Browse Source 
* feat: ensure unique artifact id

* feat: allow manual init

* fix: simplify reinit logic

* fix: no dropped value + immediate commits

* fix: wandb use in sagemaker

* docs: improve documenation and formatting

* fix: typos

* docs: improve formatting
This commit is contained in:
Boris Dayma
2021-03-22 09:45:17 -05:00
committed by GitHub
parent b230181d41
commit 125ccead71
2 changed files with 29 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
examples/README.md
View File

@@ -240,34 +240,11 @@ Whenever you use `Trainer` or `TFTrainer` classes, your losses, evaluation metri
Advanced configuration is possible by setting environment variables:
<table>
<thead>
<tr>
<th style="text-align:left">Environment Variables</th>
<th style="text-align:left">Options</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align:left">WANDB_LOG_MODEL</td>
<td style="text-align:left">Log the model as artifact at the end of training (<b>false</b> by default)</td>
</tr>
<tr>
<td style="text-align:left">WANDB_WATCH</td>
<td style="text-align:left">
<ul>
<li><b>gradients</b> (default): Log histograms of the gradients</li>
<li><b>all</b>: Log histograms of gradients and parameters</li>
<li><b>false</b>: No gradient or parameter logging</li>
</ul>
</td>
</tr>
<tr>
<td style="text-align:left">WANDB_PROJECT</td>
<td style="text-align:left">Organize runs by project</td>
</tr>
</tbody>
</table>
| Environment Variable | Value |
|---|---|
| WANDB_LOG_MODEL | Log the model as artifact (log the model as artifact at the end of training (`false` by default) |
| WANDB_WATCH | one of `gradients` (default) to log histograms of gradients, `all` to log histograms of both gradients and parameters, or `false` for no histogram logging |
| WANDB_PROJECT | Organize runs by project |
Set run names with `run_name` argument present in scripts or as part of `TrainingArguments`.