Instructions to use jinaai/xlm-roberta-flash-implementation with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use jinaai/xlm-roberta-flash-implementation with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("jinaai/xlm-roberta-flash-implementation", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Commit History
-
Fix uninitialised rotary inv_freq on transformers v5 (#63)
- Fix uninitialised rotary inv_freq on transformers v5 (44bccde5eace6e06cb5698d6c3a694a83a5d543e) Co-authored-by: Tom Aarsen <tomaarsen@users.noreply.huggingface.co>bd55a5e
-
fix: update frequencies when updating the rope base value (#40)
Verified signature- fix: update frequencies when updating the rope base value (d8cbc92c8650d6bdc8e5afb28785625a98ccfab1) - Update rotary.py (90873c4a21ac932b2df31d0e35e56b9c55460470) - Update rotary.py (071760a5bbecc7b738c64583a3b5b337cd6d0667) - Update rotary.py (1eb2361d4e9bdeedc1516196f02f199515916d30) - Update rotary.py (066b97bdf39f4031bf1ddee4c706d5c842fb8748)8b2ad1e
-
fix-device-issues (#38)
Verified signature- fix: inference issues (5365db89d3c1ad3dabfed37554c9f10e3a5918de) Co-authored-by: Jack Min Ong <Jackmin108@users.noreply.huggingface.co>57dbe22
-
Fix rotary backward
Verified signature169b7fb -
cpu-inference (#35)
Verified signature- fix cpu inference (57f6c3dc2abf60f8aa51d67551d73d5912b42146) Co-authored-by: Jack Min Ong <Jackmin108@users.noreply.huggingface.co>80b1ff7
-
refine-codebase (#33)
Verified signature- style: removing unused files, black, isort (6e55444329aa0c7f03cad24df7181b1599f8c0fc) - docs: add comments (1752c7c313018d447f01539883f352af472d8001) - configuration (3830381a980542ad592bcf3cc6c8d8cda25947fb) - chore: remove model weights (f9962ee8bc4141783cf69a239d29866bc93c5343) - docs: readme (1babf317f017ea44880d1e39a8e86ed86c087399) - chore: remove tokenizer file (37bae6d5bd9f07db0adb1e5c058490dc5c985b79) - chore: remove config (447c0db77b026b8f668214d36436c66a1cc154ce) - chore: no need for tokenizer config as well (3b9e730c66bf50cd728107c28adf0d8c2be0c207) - readme (e3fca02ce1e5e0e99c6d2cbb310a4409811de116)2646361
-
change rotary base (#31)
Verified signature- feat: rotary base as a property (c1200891411b6198ca6448cfebf5123d15bf2c31) - Merge branch 'main' into pr/31 (c2ead96805f8278295d48fda36eba1d96ed3bffb) Co-authored-by: Jack Min Ong <Jackmin108@users.noreply.huggingface.co>4434bf3
-
fix-rope-inference (#28)
Verified signature- fix: rope dim mismatch (cadf946211e1e76977f8e1e0227d657354ceb031)955fea2
-
no-flash-attention-during-inference (#22)
Verified signature- feat: no flash attention during inference (e3423c0243fe029474cbf283046f02e11e82202f)7ad815b
-
rope-embeddings (#20)
Verified signature- feat: support rope (f2e0e6205e15f5e1e23354800fe39b3ae25d9bca) - chore: remove parallelmha (8b64fa835d32d29efa8b3a34f43e4c5011ec6f13) - feat: default dim (77a17f7cb5d03e4bbfd8674fd49956fdcebb9cdc) - refactor: revert alibi stuff (11ba2000440ef53e3b8ad551ababcdf2259643ed) - chore: source (c232c27ed971d786c3e9cad242d415b2cd1fa655) - refactor: raise error if flash attention is not installed (e8e1e150c525847c940918fa1f849997e14fcaa9)e3681c2