@jzx03\n\t ,
I am not aware of any of the related code being broken right now, but Transformers is big and fast growing library, and such things happened before. It would be best if you post a reproducible example as an issue, showing difference in behavior in 4.49 and some earlier version. See how I did that in https://github.com/huggingface/transformers/issues/29525.
Also check the relevant tests like https://github.com/huggingface/transformers/blob/be37d34f44ff1bc928e59ffb8a30adecab8835a8/tests/models/llama/test_modeling_llama.py#L811 see if they still work OK, or, possibly extend the tests to cover the failure that you discovered.

\n"}}],"blogPosts":[{"_id":"6580547a45471e530841f500","isUpvotedByUser":false,"numCoauthors":0,"publishedAt":"2024-01-08T06:50:04.175Z","slug":"4d-masks","title":"4D masks support in Transformers","upvotes":31,"upvotes7d":0,"authorsData":[{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false}],"url":"/blog/poedator/4d-masks","apiPrefix":"/api/blog/poedator/4d-masks","canonical":false,"status":"published","isHFChangelog":false}],"totalBlogPosts":1,"canReadDatabase":false,"canManageEntities":false,"canReadEntities":false,"canImpersonate":false,"canManageBilling":false,"canReadRepos":false,"canReadSpaces":false,"communityScore":3,"collections":[{"slug":"poedator/nlp-664da4bb4cc4b14a8d5b6f74","title":"NLP","gating":false,"lastUpdated":"2024-05-22T07:54:35.056Z","owner":{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"items":[{"_id":"664da4bb4cc4b14a8d5b6f75","position":0,"type":"paper","id":"2405.12250","title":"Your Transformer is Secretly Linear","thumbnailUrl":"https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2405.12250.png","upvotes":157,"publishedAt":"2024-05-19T22:44:00.000Z","isUpvotedByUser":false}],"position":0,"theme":"orange","private":false,"shareUrl":"https://hf.co/collections/poedator/nlp","upvotes":0,"isUpvotedByUser":false}],"datasets":[],"models":[{"author":"poedator","authorData":{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"downloads":4177,"gated":false,"id":"poedator/opt-125m-bnb-4bit","availableInferenceProviders":[],"lastModified":"2023-11-10T15:05:31.000Z","likes":2,"pipeline_tag":"text-generation","private":false,"repoType":"model","isLikedByUser":false,"widgetOutputUrls":[],"numParameters":127942652},{"author":"poedator","authorData":{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"downloads":1,"gated":false,"id":"poedator/opt-125m-bnb-4bit_pt","availableInferenceProviders":[],"lastModified":"2023-11-10T15:04:16.000Z","likes":0,"pipeline_tag":"text-generation","private":false,"repoType":"model","isLikedByUser":false,"widgetOutputUrls":[]},{"author":"poedator","authorData":{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"downloads":1,"gated":false,"id":"poedator/open_llama_13b-gptq","availableInferenceProviders":[],"lastModified":"2023-11-09T14:38:30.000Z","likes":0,"pipeline_tag":"text-generation","private":false,"repoType":"model","isLikedByUser":false,"widgetOutputUrls":[]},{"author":"poedator","authorData":{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"downloads":1,"gated":false,"id":"poedator/open_llama_3b-gptq","availableInferenceProviders":[],"lastModified":"2023-11-09T13:35:13.000Z","likes":0,"pipeline_tag":"text-generation","private":false,"repoType":"model","isLikedByUser":false,"widgetOutputUrls":[]},{"author":"poedator","authorData":{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"downloads":0,"gated":false,"id":"poedator/classify_science_topics_TEST","availableInferenceProviders":[],"lastModified":"2022-03-27T20:42:20.000Z","likes":0,"private":false,"repoType":"model","isLikedByUser":false}],"buckets":[],"numBuckets":0,"numberLikes":2,"papers":[{"id":"2406.02532","title":"SpecExec: Massively Parallel Speculative Decoding for Interactive LLM\n Inference on Consumer Devices","thumbnailUrl":"https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2406.02532.png","upvotes":13,"publishedAt":"2024-06-04T17:53:36.000Z","isUpvotedByUser":false},{"id":"2402.12374","title":"Sequoia: Scalable, Robust, and Hardware-aware Speculative Decoding","thumbnailUrl":"https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2402.12374.png","upvotes":4,"publishedAt":"2024-02-19T18:58:32.000Z","isUpvotedByUser":false},{"id":"2306.03078","title":"SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight\n Compression","thumbnailUrl":"https://cdn-thumbnails.huggingface.co/social-thumbnails/papers/2306.03078.png","upvotes":3,"publishedAt":"2023-06-05T17:53:28.000Z","isUpvotedByUser":false}],"posts":[],"totalPosts":0,"spaces":[{"author":"poedator","authorData":{"_id":"624093c3ed76d6eadde83c48","avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","fullname":"Ruslan S.","name":"poedator","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"colorFrom":"green","colorTo":"pink","createdAt":"2022-03-27T16:59:19.000Z","emoji":"🚀","id":"poedator/scence_article_topics","lastModified":"2022-03-28T19:18:53.000Z","likes":0,"pinned":false,"private":false,"sdk":"streamlit","repoType":"space","runtime":{"stage":"RUNTIME_ERROR","hardware":{"current":null,"requested":"cpu-basic"},"storage":null,"gcTimeout":86400,"errorMessage":"Traceback (most recent call last):\n File \"/home/user/.local/bin/streamlit\", line 5, in \n from streamlit.cli import main\n File \"/home/user/.local/lib/python3.8/site-packages/streamlit/__init__.py\", line 76, in \n from streamlit.delta_generator import DeltaGenerator as _DeltaGenerator\n File \"/home/user/.local/lib/python3.8/site-packages/streamlit/delta_generator.py\", line 71, in \n from streamlit.elements.arrow_altair import ArrowAltairMixin\n File \"/home/user/.local/lib/python3.8/site-packages/streamlit/elements/arrow_altair.py\", line 25, in \n from altair.vegalite.v4.api import Chart\nModuleNotFoundError: No module named 'altair.vegalite.v4'\n","replicas":{"requested":1},"devMode":false,"domains":[{"domain":"poedator-scence-article-topics.hf.space","stage":"READY"}]},"title":"Scence_article_topics","isLikedByUser":false,"trendingScore":0,"tags":["streamlit","region:us"],"featured":false}],"u":{"avatarUrl":"/avatars/2e00e8124a1819da370101aa8e2144eb.svg","isPro":false,"fullname":"Ruslan S.","user":"poedator","orgs":[{"avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/609147936885f3a4314a6aa2/XPSp2sEWmd82vG-ztYngb.png","fullname":"Petals Team","name":"petals-team","userRole":"admin","type":"org","isHf":false,"details":"Running LLMs collaboratively"},{"avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60a551a34ecc5d054c8ad93e/Ku5nM2bKq-8ZF3Jid1ocw.png","fullname":"Blog-explorers","name":"blog-explorers","userRole":"read","type":"org","isHf":false},{"avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/636a9a07e3ad78bc68b1a5a2/J3ZMg_HOlbzVcxz9dBAgy.png","fullname":"Spec Diffusion","name":"spec-diffusion","userRole":"write","type":"org","isHf":false}],"signup":{"details":"","github":"","homepage":"","twitter":""},"isHf":false,"isMod":false,"type":"user","theme":"dark"},"upvotes":11,"numFollowers":6,"numFollowingUsers":3,"numFollowingOrgs":4,"numModels":5,"numDatasets":0,"numSpaces":1,"isFollowing":false,"isFollower":false,"sampleFollowers":[{"user":"regisss","fullname":"Régis Pierrard","type":"user","_id":"620b7c408f5871b8a1a168a7","isPro":false,"avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620b7c408f5871b8a1a168a7/49M2lucv3I24rOMJFnhVd.jpeg"},{"user":"UtopiaArtificial","fullname":"Kevin Jonathan Flores Encizo","type":"user","_id":"656077bf08eeb810e51175c5","isPro":false,"avatarUrl":"/avatars/07c407a8dba5ff9ee1c35d9c3f67c447.svg"},{"user":"bulatovv","fullname":"Albert Fazlyev","type":"user","_id":"64969ca9b4c0f362ebc26344","isPro":false,"avatarUrl":"/avatars/fcbb64ecedcf93f4fd9d713c36bec2d3.svg"},{"user":"Henrywang","fullname":"Hanlin Wang","type":"user","_id":"6261317f18176506fab13ed7","isPro":false,"avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6261317f18176506fab13ed7/PKG_OKIbrzp04bwtgAx0-.jpeg"}],"isWatching":false,"isIgnored":false,"acceptLanguages":["*"],"filters":{},"currentRepoPage":0}">