Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Appearance settings

Comments

Close side panel

Adds HF equivalency tests for Mistral#505

Open
flaviabeo wants to merge 8 commits intofoundation-model-stack:mainfoundation-model-stack/foundation-model-stack:mainfrom
flaviabeo:mistral_eq_hfflaviabeo/foundation-model-stack:mistral_eq_hfCopy head branch name to clipboard
Open

Adds HF equivalency tests for Mistral#505
flaviabeo wants to merge 8 commits intofoundation-model-stack:mainfoundation-model-stack/foundation-model-stack:mainfrom
flaviabeo:mistral_eq_hfflaviabeo/foundation-model-stack:mistral_eq_hfCopy head branch name to clipboard

Conversation

@flaviabeo
Copy link
Collaborator

No description provided.

Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
@flaviabeo flaviabeo self-assigned this Jan 27, 2026
@flaviabeo flaviabeo changed the title WIP: Adds HF equivalency tests for Mistral [WIP] Adds HF equivalency tests for Mistral Jan 27, 2026
Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
Signed-off-by: Flavia Beo <flavia.beo@ibm.com>
@flaviabeo flaviabeo marked this pull request as ready for review January 29, 2026 20:08
@flaviabeo flaviabeo changed the title [WIP] Adds HF equivalency tests for Mistral Adds HF equivalency tests for Mistral Feb 3, 2026

ratio2 = SequenceMatcher(None, output_batch[1], output_text2).ratio()

assert ratio2 > 0.8, f"text 2 incorrect - \n{output_batch[1]}\n{output_text2}"
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you for for adding these! Do the inputs currently diverge for greedy decoding? Curious about the ratio vs just matching the outputs directly

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, it's not an exact match. So assert equal always fails - These models are like a piece of the original one, with fewer heads and smaller. This way they don't produce anything so meaningful, and then when compared to the adapter one, it is not outputting fully the same texts. I thought of loosening this comparison, but please let me know your thoughts on this!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants

Morty Proxy This is a proxified and sanitized view of the page, visit original site.