Fixes to config format in documentattion

This commit is contained in:
Chime Ogbuji 2024-11-03 20:00:35 -05:00
parent 1f6c370690
commit c72122064a

View File

@ -288,19 +288,18 @@ You can specify a list of HF datasets using the `hf_datasets` (plural) configura
each with the same structure as above. For example:
```yaml
hf_datasets: [
"hf_dataset":
hf_datasets:
- hf_dataset:
name: "Open-Orca/OpenOrca"
train_split: "train[:90%]"
valid_split: "train[-10%:]"
prompt_feature: "question"
completion_feature: "response",
"hf_dataset":
completion_feature: "response"
- hf_dataset:
name: "trl-lib/ultrafeedback_binarized"
train_split: "train[:90%]"
valid_split: "train[-10%:]"
chat_feature: "chosen"
]
```
- Arguments specified in `config` will be passed as keyword arguments to