Skip to content

Commit

Permalink
Merge pull request #165 from mistralai/doc/v0.0.105
Browse files Browse the repository at this point in the history
Update docs to v0.0.105
  • Loading branch information
pandora-s-git authored Dec 4, 2024
2 parents f506abc + 3d6f984 commit 725079a
Show file tree
Hide file tree
Showing 2 changed files with 6 additions and 2 deletions.
4 changes: 4 additions & 0 deletions docs/capabilities/guardrailing.mdx
Original file line number Diff line number Diff line change
Expand Up @@ -182,6 +182,10 @@ The table below describes the types of content that can be detected in the moder
| PII | Content that requests, shares, or attempts to elicit personal identifying information such as full names, addresses, phone numbers, social security numbers, or financial account details. |


### Cookbook
Our [moderation cookbook](https://colab.research.google.com/github.com/mistralai/cookbook/tree/main/mistral/moderation/system-level-guardrails.ipynb) provides a concrete example of how to use the Moderation service to implement system level guardrails.
For a more broad view, there is also a more [explorative cookbook](https://colab.research.google.com/github.com/mistralai/cookbook/tree/main/mistral/moderation/moderation-explored.ipynb).

### FAQ
Q: What is the distribution of false-positive and false-negative results on the new moderation API models. Specifically, will they be more likely to flag something as harmful when it is not or not flag something that is harmful?

Expand Down
4 changes: 2 additions & 2 deletions docs/guides/finetuning_sections/_04_faq.md
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ For Mistral API, you can use the `auto_start=False` argument as mentioned in the

### What is the recommended learning rate?

For LoRA fine-tuning, we recommended 1e-4 (default) or 1e-5.
For LoRA fine-tuning, we recommend 1e-4 (default) or 1e-5.

Note that the learning rate we define is the peak learning rate, instead of a flat learning rate. The learning rate follows a linear warmup and cosine decay schedule. During the warmup phase, the learning rate is linearly increased from a small initial value to a larger value over a certain number of training steps. After the warmup phase, the learning rate is decayed using a cosine function.

Expand Down Expand Up @@ -99,4 +99,4 @@ for file in output_file_objects:
# now you should see three jsonl files under 500MB
```
</details>
</details>

0 comments on commit 725079a

Please sign in to comment.