Trying to handle endpoint errors

Trying to handle endpoint errors

Adding more endpoint options

Updating with new approach to inference endpoint

Endpoint name character limit

Handling of json error, running generate all at once.

Please run after fully loading

Full dataset

Clearing cached results

Backend toxicity

Moving to just toxicity

Update src/envs.py

Changing to HF_TOKEN

Changes to whatever

Changing to BACKEND_TOKEN

Toxigen

string parsing error

Fixing timing bug

changing max new tokens to None due to continuing warning on backend: Running generate_until requests: 0%| | 13/99442 [10:42<1539:30:14, 55.74s/it]Both (=2048) and (=302) seem to have been set. will take precedence. Please refer to the documentation for more information. (https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)

changing batch size to auto

evaluation with hf not hf-auto

Run full eval

Update requirements.txt

Update requirements.txt

DEBUG

DEBUG

just realtoxicityprompts

Limit to 20 now

debug

debug

debug

debug

Debug

debug

Debug

Changing refresh rate to 1 hr

Updates, not sure what -- left over from last night

Refreshing less

Changing model from hf-causal-experimental to hf-auto

Merge branch 'main' of hf.co:spaces/meg/backend into main

Trying to make it work with new EAI versions

Update requirements.txt

Trying to get code to work with an apparently diverged lm_eval

Should be realtoxicityprompts as one of the target tasks

Removing argument throwing a TypeError

Update src/backend/run_eval_suite_harness.py

Update app.py

Remove parens

Update requirements.txt

Update requirements.txt