pminervini commited on
Commit
210a90b
1 Parent(s): 67ac668
Files changed (1) hide show
  1. src/display/about.py +28 -1
src/display/about.py CHANGED
@@ -107,7 +107,34 @@ XXX
107
  """
108
 
109
  EVALUATION_QUEUE_TEXT = """
110
- XXX
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
111
  """
112
 
113
  CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
 
107
  """
108
 
109
  EVALUATION_QUEUE_TEXT = """
110
+ # Evaluation Queue for the Hallucinations Leaderboard
111
+
112
+ Models added here will be automatically evaluated on the EIDF cluster.
113
+
114
+ ## First steps before submitting a model
115
+
116
+ ### 1) Make sure you can load your model and tokenizer using AutoClasses:
117
+ ```python
118
+ from transformers import AutoConfig, AutoModel, AutoTokenizer
119
+ config = AutoConfig.from_pretrained("your model name", revision=revision)
120
+ model = AutoModel.from_pretrained("your model name", revision=revision)
121
+ tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
122
+ ```
123
+ If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded.
124
+
125
+ Note: make sure your model is public!
126
+ Note: if your model needs `use_remote_code=True`, we do not support this option yet but we are working on adding it, stay posted!
127
+
128
+ ### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index)
129
+ It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!
130
+
131
+ ### 3) Select the correct precision
132
+ Not all models are converted properly from `float16` to `bfloat16`, and selecting the wrong precision can sometimes cause evaluation error (as loading a `bf16` model in `fp16` can sometimes generate NaNs, depending on the weight range).
133
+
134
+ ## In case of model failure
135
+ If your model is displayed in the `FAILED` category, its execution stopped.
136
+ Make sure you have followed the above steps first.
137
+ If everything is done, check you can launch the EleutherAIHarness on your model locally, using the command in the About tab under "Reproducibility" with all arguments specified (you can add `--limit` to limit the number of examples per task).
138
  """
139
 
140
  CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"