Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Peteish13 #739

Merged
merged 310 commits into from
Nov 18, 2024
Merged

Peteish13 #739

merged 310 commits into from
Nov 18, 2024
This pull request is big! We’re only showing the most recent 250 commits.

Commits on Oct 4, 2024

  1. New metrics

    @dirkgr
    dirkgr committed Oct 4, 2024
    Configuration menu
    Copy the full SHA
    0656ce5 View commit details
    Browse the repository at this point in the history

  2. compile.dynamic=false

    @dirkgr
    dirkgr committed Oct 4, 2024
    Configuration menu
    Copy the full SHA
    b699753 View commit details
    Browse the repository at this point in the history

  3. Don't stop

    @dirkgr
    dirkgr committed Oct 4, 2024
    Configuration menu
    Copy the full SHA
    ca8c485 View commit details
    Browse the repository at this point in the history

Commits on Oct 5, 2024

  1. No more torch version pin

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    5f1a369 View commit details
    Browse the repository at this point in the history

  2. Run scripts to run on Augusta

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    d6cdc0c View commit details
    Browse the repository at this point in the history

  3. This runs out of memory on eval.

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    57cc09b View commit details
    Browse the repository at this point in the history

  4. Reproduce the eval we're now missing.

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    88a06bc View commit details
    Browse the repository at this point in the history

  5. Run 1000 steps

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    0f6b896 View commit details
    Browse the repository at this point in the history

  6. Let's see if dynamic=true solves the problem.

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    ce11f9f View commit details
    Browse the repository at this point in the history

  7. It did not.

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    c1d3ffe View commit details
    Browse the repository at this point in the history

  8. Set device when initializing the process group

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    d3d39a0 View commit details
    Browse the repository at this point in the history

  9. Proper way of setting a device_id

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    d8f2aac View commit details
    Browse the repository at this point in the history

  10. Maybe this incantation

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    2c0d11a View commit details
    Browse the repository at this point in the history

  11. Don't eval with a compiled model

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    6cc6f62 View commit details
    Browse the repository at this point in the history

  12. Medium LR on Weka

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    43184f3 View commit details
    Browse the repository at this point in the history

  13. Revert "Don't eval with a compiled model" 

    This reverts commit 6cc6f62.
    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    435c3e6 View commit details
    Browse the repository at this point in the history

  14. Let's try this instead.

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    a74ab6e View commit details
    Browse the repository at this point in the history

  15. Updated metrics

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    08718b9 View commit details
    Browse the repository at this point in the history

  16. This might be faster yet.

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    b093945 View commit details
    Browse the repository at this point in the history

  17. Makes the launch_train.sh script work

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    51d0a39 View commit details
    Browse the repository at this point in the history

  18. Script to run something on all nodes

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    7cf11d8 View commit details
    Browse the repository at this point in the history

  19. Back to regular compile

    @dirkgr
    dirkgr committed Oct 5, 2024
    Configuration menu
    Copy the full SHA
    212f55f View commit details
    Browse the repository at this point in the history

Commits on Oct 6, 2024

  1. new config

    @soldni
    soldni committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    68b022f View commit details
    Browse the repository at this point in the history

  2. Show all errors when something fails

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    2594a2a View commit details
    Browse the repository at this point in the history

  3. Peteish13 config for running in Google

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    4888fab View commit details
    Browse the repository at this point in the history

  4. Specify the number of nodes on the command line

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    59f2247 View commit details
    Browse the repository at this point in the history

  5. Specify the number of nodes on the command line, part 2

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    746d394 View commit details
    Browse the repository at this point in the history

  6. Correct new path of the venv

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    ea5ca3f View commit details
    Browse the repository at this point in the history

  7. Silence warning

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    b52f228 View commit details
    Browse the repository at this point in the history

  8. More informative logs

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    def6987 View commit details
    Browse the repository at this point in the history

  9. Set the number of nodes correctly

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    0b70025 View commit details
    Browse the repository at this point in the history

  10. Disable files to do a speed test

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    23f743c View commit details
    Browse the repository at this point in the history

  11. Set gcloud project in environment variable

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    a9c7daa View commit details
    Browse the repository at this point in the history

  12. New metrics

    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    a9bb215 View commit details
    Browse the repository at this point in the history

  13. Have to hold on to the GCS client so we don't overwhelm the GCS metad… 

    …ata service with credential requests
    @dirkgr
    dirkgr committed Oct 6, 2024
    Configuration menu
    Copy the full SHA
    174da42 View commit details
    Browse the repository at this point in the history

Commits on Oct 7, 2024

  1. Run with NCCL debugging

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    1678ef6 View commit details
    Browse the repository at this point in the history

  2. Merge remote-tracking branch 'origin/peteish13' into peteish13-augusta

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    fce511b View commit details
    Browse the repository at this point in the history

  3. Too much NCCL noise

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    129426a View commit details
    Browse the repository at this point in the history

  4. Merge remote-tracking branch 'origin/peteish13-augusta' into peteish1… 

    …3-augusta
    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    9858048 View commit details
    Browse the repository at this point in the history

  5. All files are there now.

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    e2a6840 View commit details
    Browse the repository at this point in the history

  6. New metrics

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    ca0abef View commit details
    Browse the repository at this point in the history

  7. Launch script for a big peteish13-medlr on Augusta

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    2f9a726 View commit details
    Browse the repository at this point in the history

  8. This will have to load from the remote save folder. No local storage … 

    …in GCP.
    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    7e8ba3f View commit details
    Browse the repository at this point in the history

  9. Kill other processes before we start

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    695f53d View commit details
    Browse the repository at this point in the history

  10. Show the output of run_all_nodes properly.

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    aba734b View commit details
    Browse the repository at this point in the history

  11. Comment out DCLM to see if this is what kills the startup

    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    7a76b03 View commit details
    Browse the repository at this point in the history

  12. Revert "Comment out DCLM to see if this is what kills the startup" 

    This reverts commit 7a76b03.
    @dirkgr
    dirkgr committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    2f94412 View commit details
    Browse the repository at this point in the history

  13. added code back

    @soldni
    soldni committed Oct 7, 2024
    Configuration menu
    Copy the full SHA
    1b020e6 View commit details
    Browse the repository at this point in the history

Commits on Oct 8, 2024

  1. Make the first host configurable

    @dirkgr
    dirkgr committed Oct 8, 2024
    Configuration menu
    Copy the full SHA
    a424b25 View commit details
    Browse the repository at this point in the history

  2. Scripts for med and high lr

    @dirkgr
    dirkgr committed Oct 8, 2024
    Configuration menu
    Copy the full SHA
    2f3bddf View commit details
    Browse the repository at this point in the history

  3. Send the WandB API key

    @dirkgr
    dirkgr committed Oct 8, 2024
    Configuration menu
    Copy the full SHA
    e66bce7 View commit details
    Browse the repository at this point in the history

  4. mask

    @soldni
    soldni committed Oct 8, 2024
    Configuration menu
    Copy the full SHA
    c40635c View commit details
    Browse the repository at this point in the history

Commits on Oct 9, 2024

  1. Enable compilation

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    054dc2d View commit details
    Browse the repository at this point in the history

  2. Peteish7 for Google

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    71284d1 View commit details
    Browse the repository at this point in the history

  3. New metrics

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    5871379 View commit details
    Browse the repository at this point in the history

  4. Make hostfiles an argument to the scripts

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    05ead0e View commit details
    Browse the repository at this point in the history

  5. Launch config for peteish7-highlr

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    e42206c View commit details
    Browse the repository at this point in the history

  6. Load the correct step0

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    c151681 View commit details
    Browse the repository at this point in the history

  7. Don;t bind to CPUs

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    eeba9cc View commit details
    Browse the repository at this point in the history

  8. More nodes

    @dirkgr
    dirkgr committed Oct 9, 2024
    Configuration menu
    Copy the full SHA
    48f7b78 View commit details
    Browse the repository at this point in the history

Commits on Oct 11, 2024

  1. Peteish7 with medium LR

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    3f7bd8f View commit details
    Browse the repository at this point in the history

  2. Don't bind to CPUs

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    2316171 View commit details
    Browse the repository at this point in the history

  3. Make the rendezvous work when the hostfile is jumbled

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    10c468b View commit details
    Browse the repository at this point in the history

  4. Enable NCCL debug

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    29b9b9c View commit details
    Browse the repository at this point in the history

  5. Use GS more

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    cbe9781 View commit details
    Browse the repository at this point in the history

  6. Shuffle nodes

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    e7c212b View commit details
    Browse the repository at this point in the history

  7. Eval every 500 steps

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    892d224 View commit details
    Browse the repository at this point in the history

  8. Eval every 500 steps

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    5315e68 View commit details
    Browse the repository at this point in the history

  9. New metrics

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    f5badaf View commit details
    Browse the repository at this point in the history

  10. Launch script to continue the low LR 13B

    @dirkgr
    dirkgr committed Oct 11, 2024
    Configuration menu
    Copy the full SHA
    ce06d18 View commit details
    Browse the repository at this point in the history

Commits on Oct 12, 2024

  1. Make the launch script work even when some nodes are commented out in… 

    … the hostsfile
    @dirkgr
    dirkgr committed Oct 12, 2024
    Configuration menu
    Copy the full SHA
    c2abab4 View commit details
    Browse the repository at this point in the history

Commits on Oct 15, 2024

  1. Adds all_reduce_bench from Stas

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    e717f44 View commit details
    Browse the repository at this point in the history

  2. Fix all_reduce_bench

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    935d695 View commit details
    Browse the repository at this point in the history

  3. Metrics interval makes a big difference

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    bc4f6b9 View commit details
    Browse the repository at this point in the history

  4. Settings for more fast peteish13

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    01669d8 View commit details
    Browse the repository at this point in the history

  5. Several jobs have made progress

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    81a2fa9 View commit details
    Browse the repository at this point in the history

  6. Script that checks node pairs for speed

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    f91cebd View commit details
    Browse the repository at this point in the history

  7. New way of launching jobs for the peteish13 config

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    6f068fa View commit details
    Browse the repository at this point in the history

  8. Experimental script to launch jobs without torchrun

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    78f9c6f View commit details
    Browse the repository at this point in the history

  9. Switch to the MPI version

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    b329bab View commit details
    Browse the repository at this point in the history

  10. Metrics

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    b15bcc4 View commit details
    Browse the repository at this point in the history

  11. Commit to hostpatterns

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    aff2124 View commit details
    Browse the repository at this point in the history

  12. run_all_nodes does not need hostpatterns, since it runs on all nodes

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    09f447c View commit details
    Browse the repository at this point in the history

  13. Forgot some more settings

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    e4a103d View commit details
    Browse the repository at this point in the history

  14. Put NCCL_DEBUG somewhere else

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    c0a130c View commit details
    Browse the repository at this point in the history

  15. Set the first host properly

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    21c8fa3 View commit details
    Browse the repository at this point in the history

  16. Reshuffle for better logs

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    f8ef856 View commit details
    Browse the repository at this point in the history

  17. Maybe high ports don't work

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    35300c0 View commit details
    Browse the repository at this point in the history

  18. Faster timeouts for debugging

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    c130118 View commit details
    Browse the repository at this point in the history

  19. First is first now

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    7f12d39 View commit details
    Browse the repository at this point in the history

  20. Merge branch 'peteish13-augusta' of https://github.com/allenai/LLM in… 

    …to peteish13-augusta
    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    a3db348 View commit details
    Browse the repository at this point in the history

  21. Revert "Faster timeouts for debugging" 

    This reverts commit c130118.
    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    3ea5a35 View commit details
    Browse the repository at this point in the history

  22. Make the launcher a little more flexible

    @dirkgr
    dirkgr committed Oct 15, 2024
    Configuration menu
    Copy the full SHA
    88ec63d View commit details
    Browse the repository at this point in the history

Commits on Oct 16, 2024

  1. new flan cleaned

    @soldni
    soldni committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    cfb7f4c View commit details
    Browse the repository at this point in the history

  2. name-fix

    @soldni
    soldni committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    19b5ad8 View commit details
    Browse the repository at this point in the history

  3. More retries for GCS

    @dirkgr
    dirkgr committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    8ff9ebd View commit details
    Browse the repository at this point in the history

  4. Metrics

    @dirkgr
    dirkgr committed Oct 16, 2024
    Configuration menu
    Copy the full SHA
    b79afee View commit details
    Browse the repository at this point in the history

Commits on Oct 17, 2024

  1. Turns out Google APIs don't work that way.

    @dirkgr
    dirkgr committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    ee08b92 View commit details
    Browse the repository at this point in the history

  2. Makes finding the latest checkpoint work on GCS

    @dirkgr
    dirkgr committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    0b3329e View commit details
    Browse the repository at this point in the history

  3. Remove unused code

    @dirkgr
    dirkgr committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    0263c7a View commit details
    Browse the repository at this point in the history

  4. Fix imports

    @dirkgr
    dirkgr committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    3e3e2e3 View commit details
    Browse the repository at this point in the history

  5. Make the code less readable

    @dirkgr
    dirkgr committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    d24a198 View commit details
    Browse the repository at this point in the history

  6. Make mypy happy

    @dirkgr
    dirkgr committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    82aa577 View commit details
    Browse the repository at this point in the history

  7. more decon

    @soldni
    soldni committed Oct 17, 2024
    Configuration menu
    Copy the full SHA
    32b869b View commit details
    Browse the repository at this point in the history

Commits on Oct 19, 2024

  1. New metrics :-(

    @dirkgr
    dirkgr committed Oct 19, 2024
    Configuration menu
    Copy the full SHA
    58cd108 View commit details
    Browse the repository at this point in the history

Commits on Oct 21, 2024

  1. decon-hard-train

    @soldni
    soldni committed Oct 21, 2024
    Configuration menu
    Copy the full SHA
    5ed69ed View commit details
    Browse the repository at this point in the history

  2. More metrics

    @dirkgr
    dirkgr committed Oct 21, 2024
    Configuration menu
    Copy the full SHA
    db18672 View commit details
    Browse the repository at this point in the history

  3. Merge commit '68b022f0bf081891704777f2d48bf836b0934d72' into peteish1… 

    …3-augusta-peteish7-annealing
    @dirkgr
    dirkgr committed Oct 21, 2024
    Configuration menu
    Copy the full SHA
    fdc998c View commit details
    Browse the repository at this point in the history

Commits on Oct 23, 2024

  1. Config for Peteish annealing

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    e90423e View commit details
    Browse the repository at this point in the history

  2. Correct path

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    4f6e96b View commit details
    Browse the repository at this point in the history

  3. Clean up launcher scripts

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    b4652d4 View commit details
    Browse the repository at this point in the history

  4. New metrics

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    2144b57 View commit details
    Browse the repository at this point in the history

  5. Annealing config for peteish7-highlr

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    ef3dca1 View commit details
    Browse the repository at this point in the history

  6. Running on half the nodes for now

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    aa84ed4 View commit details
    Browse the repository at this point in the history

  7. Merge remote-tracking branch 'origin/peteish13-augusta' into peteish1… 

    …3-augusta
    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    992b884 View commit details
    Browse the repository at this point in the history

  8. Make sure we don't train to 5T tokens

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    de3142a View commit details
    Browse the repository at this point in the history

  9. Anneal script for medium LR

    @dirkgr
    dirkgr committed Oct 23, 2024
    Configuration menu
    Copy the full SHA
    c5f2d47 View commit details
    Browse the repository at this point in the history

Commits on Oct 24, 2024

  1. Permissions

    @dirkgr
    dirkgr committed Oct 24, 2024
    Configuration menu
    Copy the full SHA
    b03f23d View commit details
    Browse the repository at this point in the history

  2. Merge branch 'peteish13-augusta' of https://github.com/allenai/LLM in… 

    …to peteish13-augusta

# Please enter a commit message to explain why this merge is necessary,
# especially if it merges an updated upstream into a topic branch.
#
# Lines starting with '#' will be ignored, and an empty message aborts
# the commit.
    @dirkgr
    dirkgr committed Oct 24, 2024
    Configuration menu
    Copy the full SHA
    b76eb1a View commit details
    Browse the repository at this point in the history

Commits on Oct 25, 2024

  1. More robust GCS downloads

    @dirkgr
    dirkgr committed Oct 25, 2024
    Configuration menu
    Copy the full SHA
    d22347c View commit details
    Browse the repository at this point in the history

  2. Metrics

    @dirkgr
    dirkgr committed Oct 25, 2024
    Configuration menu
    Copy the full SHA
    683afed View commit details
    Browse the repository at this point in the history

  3. Config to run the low LR annealing experiment

    @dirkgr
    dirkgr committed Oct 25, 2024
    Configuration menu
    Copy the full SHA
    e52cd73 View commit details
    Browse the repository at this point in the history

Commits on Oct 28, 2024

  1. Actually go to 4T

    @dirkgr
    dirkgr committed Oct 28, 2024
    Configuration menu
    Copy the full SHA
    f6c1eea View commit details
    Browse the repository at this point in the history

  2. Metrics

    @dirkgr
    dirkgr committed Oct 28, 2024
    Configuration menu
    Copy the full SHA
    05957f2 View commit details
    Browse the repository at this point in the history

Commits on Oct 29, 2024

  1. Peteish7 XHigh

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    810e23d View commit details
    Browse the repository at this point in the history

  2. Peteish7 XHigh Anneal

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    2d12eb6 View commit details
    Browse the repository at this point in the history

  3. Scripts for anneals on Beakerized Augusta

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    61bdf2b View commit details
    Browse the repository at this point in the history

  4. Fix dangerous oversight

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    9692409 View commit details
    Browse the repository at this point in the history

  5. 100B anneals

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    aa3fd74 View commit details
    Browse the repository at this point in the history

  6. Metrics

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    cf6a87f View commit details
    Browse the repository at this point in the history

  7. Second epoch for the 13B

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    37d57f9 View commit details
    Browse the repository at this point in the history

  8. Merge branch 'peteish13-augusta' of https://github.com/allenai/OLMo i… 

    …nto peteish13-augusta
    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    9015cf1 View commit details
    Browse the repository at this point in the history

  9. Inside the container, this path is different

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    76763ad View commit details
    Browse the repository at this point in the history

  10. Peteish13 anneal with unchanged data

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    e23a829 View commit details
    Browse the repository at this point in the history

  11. Merge branch 'peteish13-augusta' of https://github.com/allenai/LLM in… 

    …to peteish13-augusta
    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    51fa643 View commit details
    Browse the repository at this point in the history

  12. Need to specify epoch

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    793bf63 View commit details
    Browse the repository at this point in the history

  13. Metrics

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    ddea087 View commit details
    Browse the repository at this point in the history

  14. Leftover variable

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    11f11c8 View commit details
    Browse the repository at this point in the history

  15. One more variable

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    af2c4fd View commit details
    Browse the repository at this point in the history

  16. Ugh

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    2bb3de8 View commit details
    Browse the repository at this point in the history

  17. This is why you shouldn't write code when you're tired.

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    e69e577 View commit details
    Browse the repository at this point in the history

  18. Don't need profiles

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    9cda3de View commit details
    Browse the repository at this point in the history

  19. We should not skip this.

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    e7f3bf5 View commit details
    Browse the repository at this point in the history

  20. Install more stuff with conda

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    e056744 View commit details
    Browse the repository at this point in the history

  21. Maybe this?

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    8c710e8 View commit details
    Browse the repository at this point in the history

  22. combining flan and math

    @soldni
    soldni committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    b899c18 View commit details
    Browse the repository at this point in the history

  23. Pete says this is faster without flash anyways!

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    888a9ff View commit details
    Browse the repository at this point in the history

  24. Don't install flash

    @dirkgr
    dirkgr committed Oct 29, 2024
    Configuration menu
    Copy the full SHA
    1abccdd View commit details
    Browse the repository at this point in the history

Commits on Oct 30, 2024

  1. Launch script for 13B anneals

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    d2ca20c View commit details
    Browse the repository at this point in the history

  2. Best guess annealing config for 13B

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    ab4ceb6 View commit details
    Browse the repository at this point in the history

  3. Better mix

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    23f8f14 View commit details
    Browse the repository at this point in the history

  4. Merge remote-tracking branch 'origin/main' into peteish13-augusta 

    # Conflicts:
#	olmo/config.py
#	olmo/optim.py
#	pyproject.toml
#	scripts/lumi/log_into_node.sh
#	scripts/lumi/run-in-container.sh
#	scripts/train.py
    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    2246140 View commit details
    Browse the repository at this point in the history

  5. Better evals for the 13B anneals

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    43c8480 View commit details
    Browse the repository at this point in the history

  6. Epoch is not 1

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    9a77b60 View commit details
    Browse the repository at this point in the history

  7. Script to clear older steps

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    eb3d754 View commit details
    Browse the repository at this point in the history

  8. Fix the name, and therefore the place where we save things.

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    eafdc3e View commit details
    Browse the repository at this point in the history

  9. Merge branch 'peteish13-augusta' of https://github.com/allenai/LLM in… 

    …to peteish13-augusta
    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    8726e45 View commit details
    Browse the repository at this point in the history

  10. Eval twice as fast

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    605a3c0 View commit details
    Browse the repository at this point in the history

  11. We need FA after all.

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    4ea0d94 View commit details
    Browse the repository at this point in the history

  12. Try to load from a checkpoint

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    5408e7e View commit details
    Browse the repository at this point in the history

  13. Load latest save

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    39c964a View commit details
    Browse the repository at this point in the history

  14. We have to restore the dataloader now!

    @dirkgr
    dirkgr committed Oct 30, 2024
    Configuration menu
    Copy the full SHA
    6eec513 View commit details
    Browse the repository at this point in the history

Commits on Oct 31, 2024

  1. Peteish7 10xlr

    @dirkgr
    dirkgr committed Oct 31, 2024
    Configuration menu
    Copy the full SHA
    7ea0112 View commit details
    Browse the repository at this point in the history

  2. Continuing this run

    @dirkgr
    dirkgr committed Oct 31, 2024
    Configuration menu
    Copy the full SHA
    d7dad11 View commit details
    Browse the repository at this point in the history

Commits on Nov 1, 2024

  1. Merge commit 'b899c18fb92e14641dc84ce2a8ebf167e54d40c8' into peteish1… 

    …3-augusta
    @dirkgr
    dirkgr committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    7dba688 View commit details
    Browse the repository at this point in the history

  2. 7B anneal on Augusta/Beaker

    @dirkgr
    dirkgr committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    8031acd View commit details
    Browse the repository at this point in the history

  3. Merge branch 'peteish13-augusta' of https://github.com/allenai/LLM in… 

    …to peteish13-augusta
    @dirkgr
    dirkgr committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    7973bef View commit details
    Browse the repository at this point in the history

  4. GS authentication is a fucking joke

    @dirkgr
    dirkgr committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    c35f8c5 View commit details
    Browse the repository at this point in the history

  5. Revert "GS authentication is a fucking joke" 

    This reverts commit c35f8c5.
    @dirkgr
    dirkgr committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    ae111f4 View commit details
    Browse the repository at this point in the history

  6. Try load latest save

    @dirkgr
    dirkgr committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    c141a35 View commit details
    Browse the repository at this point in the history

  7. Set retries

    @dirkgr
    dirkgr committed Nov 1, 2024
    Configuration menu
    Copy the full SHA
    acba910 View commit details
    Browse the repository at this point in the history

Commits on Nov 2, 2024

  1. Merge remote-tracking branch 'origin/main' into peteish13-augusta

    @dirkgr
    dirkgr committed Nov 2, 2024
    Configuration menu
    Copy the full SHA
    75a1395 View commit details
    Browse the repository at this point in the history

  2. Fix merge gore

    @dirkgr
    dirkgr committed Nov 2, 2024
    Configuration menu
    Copy the full SHA
    f8f8a1e View commit details
    Browse the repository at this point in the history

  3. So urgent!

    @dirkgr
    dirkgr committed Nov 2, 2024
    Configuration menu
    Copy the full SHA
    4f5866b View commit details
    Browse the repository at this point in the history

  4. Don't wait so long for startup

    @dirkgr
    dirkgr committed Nov 2, 2024
    Configuration menu
    Copy the full SHA
    21edc84 View commit details
    Browse the repository at this point in the history

  5. Don't overwrite checkpoints!

    @dirkgr
    dirkgr committed Nov 2, 2024
    Configuration menu
    Copy the full SHA
    2e1fd4a View commit details
    Browse the repository at this point in the history

  6. Cleanup of this config

    @dirkgr
    dirkgr committed Nov 2, 2024
    Configuration menu
    Copy the full SHA
    a1271eb View commit details
    Browse the repository at this point in the history

Commits on Nov 3, 2024

  1. Consolidate the 1B config

    @dirkgr
    dirkgr committed Nov 3, 2024
    Configuration menu
    Copy the full SHA
    cb9c22a View commit details
    Browse the repository at this point in the history

  2. Higher priority

    @dirkgr
    dirkgr committed Nov 3, 2024
    Configuration menu
    Copy the full SHA
    acf1ccc View commit details
    Browse the repository at this point in the history

  3. Smaller timeout

    @dirkgr
    dirkgr committed Nov 3, 2024
    Configuration menu
    Copy the full SHA
    3de3c8e View commit details
    Browse the repository at this point in the history

  4. Let's just all run with high so we don't preempt each other.

    @dirkgr
    dirkgr committed Nov 3, 2024
    Configuration menu
    Copy the full SHA
    e598e0c View commit details
    Browse the repository at this point in the history

  5. Peteish1 configs

    @dirkgr
    dirkgr committed Nov 3, 2024
    Configuration menu
    Copy the full SHA
    cb740a6 View commit details
    Browse the repository at this point in the history

Commits on Nov 4, 2024

  1. More data loading workers

    @dirkgr
    dirkgr committed Nov 4, 2024
    Configuration menu
    Copy the full SHA
    74e30ca View commit details
    Browse the repository at this point in the history

  2. Change peteish7 lr anneals so that they resume from the middle

    @dirkgr
    dirkgr committed Nov 4, 2024
    Configuration menu
    Copy the full SHA
    de19d65 View commit details
    Browse the repository at this point in the history

  3. With retries!

    @dirkgr
    dirkgr committed Nov 4, 2024
    Configuration menu
    Copy the full SHA
    a6b1c36 View commit details
    Browse the repository at this point in the history

  4. Need to specify this now :-(

    @dirkgr
    dirkgr committed Nov 4, 2024
    Configuration menu
    Copy the full SHA
    bdb6c13 View commit details
    Browse the repository at this point in the history

  5. Bounce epoch

    @dirkgr
    dirkgr committed Nov 4, 2024
    Configuration menu
    Copy the full SHA
    1ac0437 View commit details
    Browse the repository at this point in the history

Commits on Nov 5, 2024

  1. Settings for speed!

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    e87265c View commit details
    Browse the repository at this point in the history

  2. We are urgently important.

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    02ae870 View commit details
    Browse the repository at this point in the history

  3. We are only highly important.

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    62927bd View commit details
    Browse the repository at this point in the history

  4. Another config from Pete

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    0e8246e View commit details
    Browse the repository at this point in the history

  5. Try making more GPUs warm

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    8b9b0e4 View commit details
    Browse the repository at this point in the history

  6. Don't eval so often

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    9b1c670 View commit details
    Browse the repository at this point in the history

  7. Even less frequently

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    f33abac View commit details
    Browse the repository at this point in the history

  8. High LR config for the 1B

    @dirkgr
    dirkgr committed Nov 5, 2024
    Configuration menu
    Copy the full SHA
    cdf4319 View commit details
    Browse the repository at this point in the history

Commits on Nov 6, 2024

  1. muP LR

    @dirkgr
    dirkgr committed Nov 6, 2024
    Configuration menu
    Copy the full SHA
    6ca55e7 View commit details
    Browse the repository at this point in the history

Commits on Nov 7, 2024

  1. Merge remote-tracking branch 'origin/main' into peteish13-augusta 

    # Conflicts:
#	olmo/util.py
    @dirkgr
    dirkgr committed Nov 7, 2024
    Configuration menu
    Copy the full SHA
    ede2464 View commit details
    Browse the repository at this point in the history

  2. Anneal with different random seeds

    @dirkgr
    dirkgr committed Nov 7, 2024
    Configuration menu
    Copy the full SHA
    2ba7632 View commit details
    Browse the repository at this point in the history

  3. Fix settings

    @dirkgr
    dirkgr committed Nov 7, 2024
    Configuration menu
    Copy the full SHA
    57dac1c View commit details
    Browse the repository at this point in the history

  4. Peteish7 medlr on Beaker

    @dirkgr
    dirkgr committed Nov 7, 2024
    Configuration menu
    Copy the full SHA
    d78fb17 View commit details
    Browse the repository at this point in the history

  5. Forgot to substitute variable

    @dirkgr
    dirkgr committed Nov 7, 2024
    Configuration menu
    Copy the full SHA
    2ee05eb View commit details
    Browse the repository at this point in the history

  6. Need different settings for Augusta

    @dirkgr
    dirkgr committed Nov 7, 2024
    Configuration menu
    Copy the full SHA
    6819484 View commit details
    Browse the repository at this point in the history

  7. We can get away with this.

    @dirkgr
    dirkgr committed Nov 7, 2024
    Configuration menu
    Copy the full SHA
    e688f07 View commit details
    Browse the repository at this point in the history

Commits on Nov 8, 2024

  1. Peteish13 on Beaker

    @dirkgr
    dirkgr committed Nov 8, 2024
    Configuration menu
    Copy the full SHA
    1959960 View commit details
    Browse the repository at this point in the history

  2. Fix variable

    @dirkgr
    dirkgr committed Nov 8, 2024
    Configuration menu
    Copy the full SHA
    a489d84 View commit details
    Browse the repository at this point in the history

  3. Set the remote folder sensibly

    @dirkgr
    dirkgr committed Nov 8, 2024
    Configuration menu
    Copy the full SHA
    21143d4 View commit details
    Browse the repository at this point in the history

  4. Forgot to put the seed in the run name

    @dirkgr
    dirkgr committed Nov 8, 2024
    Configuration menu
    Copy the full SHA
    cdc544c View commit details
    Browse the repository at this point in the history

  5. Wrangle the variables some more

    @dirkgr
    dirkgr committed Nov 8, 2024
    Configuration menu
    Copy the full SHA
    76e42d7 View commit details
    Browse the repository at this point in the history

Commits on Nov 9, 2024

  1. Fix wandb

    @dirkgr
    dirkgr committed Nov 9, 2024
    Configuration menu
    Copy the full SHA
    df55ddb View commit details
    Browse the repository at this point in the history

  2. Peteish13 on Beaker

    @dirkgr
    dirkgr committed Nov 9, 2024
    Configuration menu
    Copy the full SHA
    3dbc034 View commit details
    Browse the repository at this point in the history

  3. Another annealing config

    @dirkgr
    dirkgr committed Nov 9, 2024
    Configuration menu
    Copy the full SHA
    dbaa442 View commit details
    Browse the repository at this point in the history

  4. Anneal for peteish7 medlr

    @dirkgr
    dirkgr committed Nov 9, 2024
    Configuration menu
    Copy the full SHA
    2e635cf View commit details
    Browse the repository at this point in the history

  5. Actually run to 5T

    @dirkgr
    dirkgr committed Nov 9, 2024
    Configuration menu
    Copy the full SHA
    842ad6a View commit details
    Browse the repository at this point in the history

Commits on Nov 11, 2024

  1. Try with flash

    @dirkgr
    dirkgr committed Nov 11, 2024
    Configuration menu
    Copy the full SHA
    f18381e View commit details
    Browse the repository at this point in the history

Commits on Nov 12, 2024

  1. Dockerfile

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    3a2d62e View commit details
    Browse the repository at this point in the history

  2. Change image

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    0e7b088 View commit details
    Browse the repository at this point in the history

  3. Can't have retries when it's configured like this.

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    320fd7b View commit details
    Browse the repository at this point in the history

  4. So urgent!

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    6099826 View commit details
    Browse the repository at this point in the history

  5. Document what we're doing.

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    30eb3a6 View commit details
    Browse the repository at this point in the history

  6. We can retry again!

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    f6c4cb8 View commit details
    Browse the repository at this point in the history

  7. Run peteish1 on fewer nodes

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    2ec3fa0 View commit details
    Browse the repository at this point in the history

  8. Actually use all the nodes

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    17b2bca View commit details
    Browse the repository at this point in the history

  9. No retries while we're debugging

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    eba9075 View commit details
    Browse the repository at this point in the history

  10. Try longer to start up

    @dirkgr
    dirkgr committed Nov 12, 2024
    Configuration menu
    Copy the full SHA
    7a35ce9 View commit details
    Browse the repository at this point in the history

Commits on Nov 13, 2024

  1. Anneal the anneal

    @dirkgr
    dirkgr committed Nov 13, 2024
    Configuration menu
    Copy the full SHA
    34d12fe View commit details
    Browse the repository at this point in the history

  2. Annealing config for the 1B

    @dirkgr
    dirkgr committed Nov 13, 2024
    Configuration menu
    Copy the full SHA
    8908d65 View commit details
    Browse the repository at this point in the history

  3. Tix fypo

    @dirkgr
    dirkgr committed Nov 13, 2024
    Configuration menu
    Copy the full SHA
    c5b0369 View commit details
    Browse the repository at this point in the history

  4. Turn flash attention back on for 7B anneals

    @dirkgr
    dirkgr committed Nov 13, 2024
    Configuration menu
    Copy the full SHA
    b3668f4 View commit details
    Browse the repository at this point in the history

  5. No whammy 3 config

    Matt Jordan committed Nov 13, 2024
    Configuration menu
    Copy the full SHA
    d657a19 View commit details
    Browse the repository at this point in the history

  6. Urgent

    @dirkgr
    dirkgr committed Nov 13, 2024
    Configuration menu
    Copy the full SHA
    35aaaf3 View commit details
    Browse the repository at this point in the history

  7. Merge branch 'peteish13-augusta' of https://github.com/allenai/LLM in… 

    …to peteish13-augusta
    @dirkgr
    dirkgr committed Nov 13, 2024
    Configuration menu
    Copy the full SHA
    7dc73b3 View commit details
    Browse the repository at this point in the history

Commits on Nov 14, 2024

  1. added config

    @soldni
    soldni committed Nov 14, 2024
    Configuration menu
    Copy the full SHA
    bca921f View commit details
    Browse the repository at this point in the history

Commits on Nov 15, 2024

  1. Run with retries

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    7b0c0f8 View commit details
    Browse the repository at this point in the history

  2. Don't run out of space.

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    949d80c View commit details
    Browse the repository at this point in the history

  3. Merge branch 'peteish13-augusta' of https://github.com/allenai/LLM in… 

    …to peteish13-augusta
    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    ab7d870 View commit details
    Browse the repository at this point in the history

  4. Remove all Augusta specific stuff

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    93df396 View commit details
    Browse the repository at this point in the history

  5. Fix paths

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    bad96cd View commit details
    Browse the repository at this point in the history

  6. Remove unused config

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    214aea5 View commit details
    Browse the repository at this point in the history

  7. Delete all the LUMI scripts

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    200bd1f View commit details
    Browse the repository at this point in the history

  8. Remove metrics notebook

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    5d8da46 View commit details
    Browse the repository at this point in the history

  9. Changelog

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    8b709b9 View commit details
    Browse the repository at this point in the history

  10. Productivity through formatting

    @dirkgr
    dirkgr committed Nov 15, 2024
    Configuration menu
    Copy the full SHA
    c7c0c5b View commit details
    Browse the repository at this point in the history

Commits on Nov 16, 2024

  1. Config for more 13B anneals

    @dirkgr
    dirkgr committed Nov 16, 2024
    Configuration menu
    Copy the full SHA
    6f4a49a View commit details
    Browse the repository at this point in the history

  2. Needs to be urgent

    @dirkgr
    dirkgr committed Nov 16, 2024
    Configuration menu
    Copy the full SHA
    4e57631 View commit details
    Browse the repository at this point in the history

Commits on Nov 17, 2024

  1. Hopefully fix a weird warning?

    @dirkgr
    dirkgr committed Nov 17, 2024
    Configuration menu
    Copy the full SHA
    18b2f6a View commit details
    Browse the repository at this point in the history

  2. Anneal Dirk's checkpoint

    @dirkgr
    dirkgr committed Nov 17, 2024
    Configuration menu
    Copy the full SHA
    0b37469 View commit details
    Browse the repository at this point in the history

  3. That never should have said "-google".

    @dirkgr
    dirkgr committed Nov 17, 2024
    Configuration menu
    Copy the full SHA
    2997b24 View commit details
    Browse the repository at this point in the history

  4. Two more anneals for the queue

    @dirkgr
    dirkgr committed Nov 17, 2024
    Configuration menu
    Copy the full SHA
    c0fc15e View commit details
    Browse the repository at this point in the history

Commits on Nov 18, 2024

  1. Same but longer

    @dirkgr
    dirkgr committed Nov 18, 2024
    Configuration menu
    Copy the full SHA
    96e4b0d View commit details
    Browse the repository at this point in the history

  2. Two more configs

    @dirkgr
    dirkgr committed Nov 18, 2024
    Configuration menu
    Copy the full SHA
    a0c09ea View commit details
    Browse the repository at this point in the history