Spaces:
Running on Zero
Running on Zero
Commit History
3 days ago
-
Download in the worker, and declare the format Running
Verified signatureDownloads the beat checkpoint synchronously in the worker, moves the model card first, and declares 32 kHz mono. It also imports spaces plainly, since it is always installed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>f71a5c5
5 days ago
-
Pin pyharp v0.4.2.
Verified signature6b44d88
7 days ago
-
Build the models at startup, check inputs before reserving the GPU, bound the retries, and reserve 170 s rather than 300.
Verified signature795f903
11 days ago
-
Finish loading in pyharp's worker before process_fn can reserve a GPU.
Verified signature72ad407 -
Wait for the worker's own download thread instead of failing the first request.
Verified signature889a592 -
Pin pyharp v0.4.1 by commit, since pip caches the wheel it built for a moved tag.
Verified signature5dfbd6f
13 days ago
-
Pin pyharp by commit, since pip caches the wheel it built for a moved tag.
Verified signature59ae679
14 days ago
-
Tidy the launch call and drop stray files
Verified signaturef3ba3f4 -
Roll pyharp and tidy the card description
Verified signatureab81a13 -
Prune tags that predate the taxonomy
Verified signature85e7ce5 -
Read the beat checkpoint from the weights repository, not from another Space
Verified signature2545498
15 days ago
-
Declare the more restrictive of the code and weights licenses
Verified signaturecbff866 -
Declare the more restrictive of the code and weights licenses
Verified signature7eb9512 -
Tidy the description's formatting
Verified signature9ab0a3a -
Declare the code's license and state the weights' terms in the description
Verified signaturea187165 -
Credit the original authors and link the repository and paper
Verified signature2aa88e1 -
Record the model's tags, as reported by /controls
Verified signature16f58b9 -
Fix the launch keywords and the taxonomy import
Verified signature4a6e1a6 -
Bring the Space up to the pyharp example standard (taxonomy tags, __main__ guard, show_error, v0.4.0)
Verified signature5447224 -
Correct the license and give the Space an apt emoji and gradient
Verified signature7446af0
16 days ago
-
Bring Space metadata up to HARP deployment guidelines
Verified signaturec95a039
-
cleanup: remove dead MAGNeT/MBD loader code, drop redundant dbn=False, remove loopgen-repo reference copy
Vansh Chugha80fd99 -
pin dependencies
Vansh Chugh39c9833 -
force fresh requirements install for build-log pinning
Vansh Chughc1680eb -
remove debug prints, restore gpu duration
Vansh Chughf8c94d1 -
note in model card that output loops twice
Vansh Chughfffb8ef -
add debug prints for loop length investigation
Vansh Chugh29294cb -
match original pipeline's loop-doubling at max duration
Vansh Chughc0089ac -
shorten comment
Vansh Chugh828d286 -
fix mask padding trim missing from non-memory-efficient attention path
Vansh Chugh6f9efe3 -
route magnet through custom attention path instead of nn.MultiheadAttention
Vansh Chughafb6c50 -
revert dead-code edit in loaders.py, real fix is in loopgen.py
Vansh Chughfeb804f -
disable memory-efficient attention in loopgen's own magnet loader
Vansh Chughf34b253 -
disable memory-efficient attention for magnet
Vansh Chughc875b65
-
try torch 2.10.0 to match xformers build, shrink gpu duration for testing
Vansh Chugh9da6de5 -
disable memory-efficient attention for musicgen too
Vansh Chugh3ebe4e2 -
disable memory-efficient attention, xformers has no blackwell kernel
Vansh Chughdf3d717 -
bump torch/xformers to fix flash-attn crash on zerogpu
Vansh Chugh77b301e -
initial deploy
Vansh Chugh83e59db -
initial commit
Verified signature0b9c252