Loading
1 Follower
1 Following
kimberley_jensen

Location

GB

Badges

2
0
0

Connect

Activity

Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Mon
Wed
Fri

Ratings Progression

Loading...

Challenge Categories

Loading...

Challenges Entered

Latest submissions

See All
graded 220393
failed 220379
graded 220364

Latest submissions

See All
graded 165257
failed 165224
failed 165087

Music source separation of an audio signal into separate tracks for vocals, bass, drums, and other

Latest submissions

See All
graded 220393
failed 220379
graded 220364

Source separation of a cinematic audio track into dialogue, sound-effects and misc.

Latest submissions

No submissions made in this challenge.
Participant Rating
alina_porechina 0
Participant Rating
alina_porechina 0

Sound Demixing Challenge 2023

Clean version of Moises SDXDB23

2 months ago

(post deleted by author)

Clean version of Moises SDXDB23

4 months ago

Please can we have access to the clean version of the Moises SDXDB23 that was used for Leaderboard A and B

MDX Leaderboard C 4th place submission

4 months ago

GitHub - KimberleyJensen/kmdx-net_music-source-separation - Submission code with models download

Model Summary

  • Models
  • Demucs models (all trained by meta)
    • 4 x htdemucs_ft
    • demucs_mmi
    • a1d90b5c
  • Kuielab mdx-net models (all trained by me on a single 16gb T4 GPU)
    • vocals.onnx (trained with batchnorm2d, the vocals model is an onnx file because it was trained before i realized pytorch models were faster and i deleted the checkpoint)
    • bass.pt (trained with groupnorm num_groups=4)
    • other.pt (trained with groupnorm num_groups=2)
    • drums.pt (trained with groupnorm num_groups=2)
  • Things i did to boost SDR
    • modify the input mixtures for each model
      • input mixture for htdemucs_ft vocals = The original mixture
      • input mixture for htdemucs_ft drums = Original mixture - output of htdemucs_ft vocals
      • input mixture for htdemucs_ft bass = Original mixture - output of htdemucs_ft vocals + drums
      • input mixture for bass, drums and other.pt = Original mixture - output of vocals.onnx
    • Using original mixture - output of htdemucs_ft vocals + drums + bass as the other stem scored higher than using the actual other stem model.
    • Blending the model outputs
      • First i blended the demucs model outputs together, then blended the demucs outputs with the mdx-net model outputs
      • [0.08, 0.08, 0.4, 0.88] was the final blend value between demucs and mdx-net (0 would mean 100% demucs was used and 1 would mean 100% mdx-net) in the order of drums,bass, other and vocals.

Dataset details

For the vocals.onnx model the dataset contains 2000 acapellas and instrumentals

bass.pt was Musdb18 + 200 extra bass and bassless tracks

drums.pt was Musdb18 + 150 extra drums and drumless tracks

other.pt was Musdb18 + 100 extra other and otherless tracks.

Training code for kuielab mdx-net with modified model settings GitHub - KimberleyJensen/mdx-net at mdx23

Aicrowd_gym ModuleNotFoundError

5 months ago

@crlandsc

When i had the issue i just forked a new page from the starter kit made the changes for my code and models and then it worked. AIcrowd / Challenges / Sound Demixing Challenge 2023 / SDX 2023 Music Demixing Track Starter Kit · GitLab

Submissions are being stuck on "Preparing the cluster for you" for a long time

5 months ago

@mohanty

Thank you for your reply, I have not been able to make any of my final submissions that get evaluated for 4 days and now there is only 2 days left, my final submissions would improve my scores and position on the leaderboard and it could be the same for other people in the contest.

I think this challenge should be paused or extended by the amount of time since the issue started until the issue is fixed.

Submissions are being stuck on "Preparing the cluster for you" for a long time

5 months ago

@dipam @mohanty @StefanUhlich @GiorgioFabbro

Please can this be looked into there is only 3 days left of this challenge and people havent been able to make there final submissions, i tried contacting you by email help@aicrowd.com but thats broken aswell

Submissions are being stuck on "Preparing the cluster for you" for a long time

5 months ago

@dipam @mohanty @snehananavati

My submissions are getting stuck on the Preparing the cluster for you phase and it looks like everyone elses submissions are too, some submissions have been submitted for 8+ hours and havent been graded.

Submissions failing

6 months ago

@dipam @mohanty

Please can you check if there is any issues for when the inference timer begins on the main run? Because my inference speed on validation is 2.103x and the allowed time is 1.68x, there is nothing in my code that will cause that much difference, i have run lots of local evaluations and the only difference is like 3 seconds between evaluations.

Is it normal when going between validation and the main run to have “Preparing the cluster for you
PodInitializing” and “PodInitializing” happen again?

Also is there any better way to talk about my issue? Like on discord? Because only getting one reply per 1-2 days is not good for me when the contest ends in 20 days

Submissions failing

6 months ago

@dipam

Do you have any suggestions why the inference in my submissions for validation works but for the main run it fails please? These are debug logs

submission: AIcrowd

debug logs: AIcrowd
admin logs: https://dashboard.aicrowd.com/explore?orgId=15&left={"datasource":"Logs","queries":[{"refId":"A","expr":"{aicrowd_com_submission_id%3D\"214465\",%20aicrowd_com_logs%3D\"true\",%20container%3D\"main\"}"}],"range":{"from":"now-2d","to":"now"}}

Submissions failing

6 months ago

@dipam @mohanty

My issue is being caused by pytorch 2.0.0, my code requires pytorch 1.13.1 and I have it listed in my requirements.txt as torch==1.13.1 but it seems that the Music Demixing Validation and Music Demixing are using different environments?

Please can you tell me how to fix this error?

Submissions failing

6 months ago

@dipam @mohanty

I dont think this is caused by submission timeout because the Music Demixing Validation part would have failed aswell

Submissions failing

6 months ago

Hello my submissions have started randomly failing, I complete the Music Demixing Validation but then fail on the Music Demixing

Link to submission AIcrowd

Here are the debug logs AIcrowd

and the admin reference logs https://dashboard.aicrowd.com/explore?orgId=15&left={"datasource":"Logs","queries":[{"refId":"A","expr":"{aicrowd_com_submission_id%3D\"213912\",%20aicrowd_com_logs%3D\"true\",%20container%3D\"main\"}"}],"range":{"from":"now-2d","to":"now"}}

Submission failed : Failed to communicate with the grader

8 months ago

Some of my submissions are getting the error “Submission failed : Failed to communicate with the grader.”

People from another challenge are getting the error aswell

Evaluation failed due to system error

9 months ago

@mohanty

I think there is a problem with the computers you are using for evaluation because i had this problem with one of my evaluations and no problems with another without changing any code. Please could you check the PCs you use because some of them are performing a lot slower than others

Feedback and suggestion

9 months ago

I am having the same problem with throughput. These two pictures use the same models but the second one took 3 times longer and timed out


Music Demixing Challenge ISMIR 2021

Ask your questions on the townhall presentation for Improved Demucs here

About 2 years ago

Hi @defossez :slightly_smiling_face: It’s ok if you cant answer all of these questions :laughing:

How much extra data did you use for leaderboard B?

Could you make Demucs better if there was a longer time limit for predictions?

Did you use model blending for your submissions?

Are you going to keep working on Demucs to try and make it even better?

When can we train our own models on the new Demucs and when can we read your paper about it?

Thanks !

🎉 The final leaderboard is live

About 2 years ago

Well done to the winners and everyone else !!

Please would it be possible to have access to the dataset we were testing against? I want to try and keep improving my model and seeing how high i can get my score. And are we able to listen to our separations for all the songs?

PS Thanks to @woosung_choi and @defossez they both helped me a lot by answering my questions and helping me get started :slightly_smiling_face: :slightly_smiling_face:

The leaf - wicked drums file from musdb18 has the electric guitar playing on it

About 2 years ago

i just found out that the drums file from the musdb18 song Leaf - Wicked has the electric guitar on it. The electric guitar starts at 2:20

i will share if i find anymore

My submission failed because of AssertionError:

About 2 years ago

Thank you !! I just got my first submission graded

kimberley_jensen has not provided any information yet.