Spaces:
Running on CPU Upgrade
Running on CPU Upgrade
Commit History
-
Don't force qwen-asr hub<1.0 stack on native Qwen3-ASR models
Shivam +1: CursorModel ids containing "qwen3-asr" were matched by name and pinned to huggingface-hub<1.0 (qwen-asr / transformers==4.57.6). Native `-hf` checkpoints use the Transformers 5.13 port and need hub>=1.0, so the job failed to resolve deps. - Detect the native format from the Hub config (no thinker_config + flat text_config) and skip the qwen-asr stack for it. - Let an explicit transformers>=5.x requirement override the model-id and recipe heuristics. - Add load_hub_config() and refactor infer_model_type() onto it. Co-authored-by: Cursor <cursoragent@cursor.com>a176818
-
model param fetch, text edit
Shivam65f8690
-
granite 512 tokens again
b26adce -
changing max tokens for granite
bd1dce8 -
granite debug
0eab562
-
Revert "cleaning moderate section"
This reverts commit 0377732d72f547a1385a24a68ff4143953ec2706.35741ac -
cleaning moderate section
0377732 -
custom eval fix
7360365
-
edits to custom installations
b692e9f -
custom script support
83a4b95 -
added moving source eval
0ac9178
-
pyannote add
44734af -
cohere fix
a6beab2
-
custom eval update
a65ea12 -
cleanup
c6a8e55 -
beautification, custom scripts loader, artefacts json to csv
c0dd455 -
more requirement update
15eefec -
backend dependent requirements
881394f -
frontend fixes
958572a
-
backend fix
91158b9
-
multiple changes including the local eval
Shivamb5211b9
-
Visualization fixes
Shivam751af19 -
bug fixes
Shivam6d689d0 -
evaluation methods update
Shivam63a45c6
-
code fixes
Shivame9b0a98 -
error fix
Shivam19c0dc7 -
added support for remote models
Shivam1d9cf7f -
New leaderboard method updating script per class
Shivam66dddc0 -
Refactor ASR evaluation into pluggable backends and add model family dropdown.
ShivamMade-with: Cursor8a47521