Spaces:
Runtime error
Runtime error
Commit History
-
Replace hardcoded architecture detection with user selection
🎯 Key Changes: - Remove hardcoded architecture patterns (future-proof for new models) - Add model type dropdown to submission form: 🔤 🔽 🔄 - Users now select: Encoder/Decoder/Encoder-Decoder during submission - Model type stored in request files and displayed in T column ✨ Benefits: - Works with any new/custom architecture - User explicitly declares their model type - No need to maintain hardcoded pattern lists - Clear UI with emoji + description 🔤 Encoder (BERT-like) < /dev/null | 🔽 Decoder (GPT-like) | 🔄 Encoder-Decoder (T5-like) 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>17f029a -
Fix leaderboard startup and simplify to core functionality
Major fixes: - Fix fields() function to work with both make_dataclass and @dataclass - Fix column name mapping (model -> Model, average -> Average, etc.) - Fix JSON file filtering logic that was skipping result files - Fix search_columns references to use correct case-sensitive names - Remove unnecessary metadata columns (precision, license, params, etc.) - Simplify to core leaderboard: Model name + task scores + average The app now starts successfully and displays a clean leaderboard focused on model performance comparison across NER tasks. 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>25c6939 -
Fix critical stability issues in leaderboard
- Fix KeyError when tasks missing from evaluation results - Prevent division by zero in average calculations - Add safe DataFrame column access in filtering - Fix file path bugs in subdirectory processing - Add JSON error handling for malformed queue files - Improve license access with proper fallbacks - Make filter columns dynamic based on available data These fixes prevent major crash scenarios when handling edge cases, empty data, or malformed evaluation files. 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>9f4fde3 -
fix
1e26b4a -
Simplify leaderboard to EMEA-sen and MEDLINE tasks only
- Update Tasks enum to only include emea_ner and medline_ner - Simplify interface text and remove unnecessary complexity - Update French localization - Remove hardcoded dataset information 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>fefe31a
-
Handle empty DataFrame case in leaderboard
ad128a8 -
Add fallback sorting when no task results exist
a18de40 -
Fix PyTorch dependency and Average column KeyError
dbd8502 -
Add French medical NER leaderboard frontend
81722bf
-
Apply CSS to the model name column (#14)
- Apply CSS to the model name column (e5f0365777f9f167164b69a0afcb561fe083b075) Co-authored-by: Takumi Okamoto <t0-0@users.noreply.huggingface.co>fc32710
-
os.isfile --> os.path.isfile (#12)
- os.isfile --> os.path.isfile (5b12d1ec95fbe51d25ef7875807960b493df2a46) Co-authored-by: Margaret Mitchell <meg@users.noreply.huggingface.co>abf5a88
-
Minimal edit for directory structure assumption. (#11)
- Minimal edit for directory structure assumption. (6513f2eccd9fc6e003944c90c2ada29a2576a10b) Co-authored-by: Margaret Mitchell <meg@users.noreply.huggingface.co>1e9865f
-
Update src/display/utils.py
f867bb5
-
Update src/envs.py
30da67d
-
added leaderboard component to simplify main script
Clémentine4103566
-
doc
Clémentine5fe3b95 -
simplified the template
Clémentinebbd72ab
-
CPU, TOKEN, env variables (#4)
- Suggesting change for accidentally hard-coded environment variable (8d208b520eea9413d6513b14ce2198e0fe0dca18) - Adds CPU/Float32 support; adds user access Token passing; fixes hard-coded environment variable. (53820335700fc97ce312f7f9fbc7a36325e0aef9) - Just trying to open a PR without access to the org through the command line. (df50a1a673c11d256fdc3d782644a9b1fea0c27a) - Creating branch of changes through command line. (d3ce128251987516e6d2c37d6c2c85db6e9a7934) - Test (90f11e5b611eb9474f877b99c832a5bc7c30f59a) - Remove tmp file (67d160e129e6b3ccdb2cb54ba6c161ce82ff7501) - Undo 1 change; keeping 'eval-queue' hard-coded. (5866f269eb67d1d65c2839651bf7be1ff7e245a6) - Undo 1 change; keeping 'eval-queue' hard-coded. (1e36a73ac445fc2cc60200b39a5f31361107a283) Co-authored-by: Margaret Mitchell <meg@users.noreply.huggingface.co>4d3bd61
-
Update src/submission/check_validity.py
0767925
-
made token a requirement
Clémentine0df3db8 -
test
Clémentine16837ad
-
fix
Clémentine6be74f5 -
removed quantization to simplify
Clémentine32b707a -
now with a functionning backend
Clémentine01ea22b
-
update read
Clémentine4f3c2a8 -
fixs
Clémentine91e8a06 -
updated leaderboard
Clémentine29546b4 -
Simplified leaderboard v0
Clémentine2a860f6 -
simplified some parts of the code + updated requirements
Clémentine3d8dbe8 -
Added check on tokenizer to prevent submissions which won't run
Clémentine06b0c2d -
Update benchmark count and fix typo (`inetuning->finetuning`) (#395)
- Update benchmark count and fix typo (`inetuning->finetuning`) (cdeea55b7621c0b1fa7515a40bf2fb50df62d5d7) Co-authored-by: Alvaro Bartolome <alvarobartt@users.noreply.huggingface.co>71f3e62
-
fix order of request file vs request file list, to avoid resubmitting issues
Clémentine54eae7e -
cache
Clémentinea983535 -
update for caching
Clémentine29efd58 -
add model architecture as column
Clémentine1ba1924 -
Simplify About
Clémentinea6eaff8
-
Refactor 2 - added plotting back
ClémentineOnly takes into account last submissions, but we have no way to go back at eval date apart from loading info from git commit of results files. Also updated speed with gradio concurrency limit paramceb2102 -
fix value error in param size
Clémentine76a47ce -
Fix requirements for mistral models - to change once transformers gets updated.
ClémentineWe need hf_hub >= 0.18 for collections, transformers >= 4.34 for mistral, and both depend on uncompatible versions of tokenizers. Installing transformers from a fork for nowc875275
-
fix col width
Clémentine16a06c4 -
refacto style + rate limit
Clémentine3b3db42 -
Fix TruthfulQA NaN scores to 0
Clémentine992b592
-
refacto part 1
Clémentine8b1f7a0
-
add new evals to the leaderboard
Nathan Habib2a73df7
-
token for checking gated base models
Clémentine8f3a3ec
-
Fix BibTex author ordering (#342)
- Fix BibTex author ordering (2edeb5ab568075636e23b5ecfd11c297c4425330) Co-authored-by: Lewis Tunstall <lewtun@users.noreply.huggingface.co>bdabf34 -
fix disapearing models
Nathan Habib3817014
-
Merge branch 'main' of https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard
Nathan Habibcc5299c -
fix model display when fething metadata
Nathan Habib95eba41
-
reorg to simplify nav in code base
Clémentinec0da33d