[peculator training] Update benchmark_speculator_logical.py to support gpt_bigcode/granite#62
Open
sahilsuneja1 wants to merge 10 commits intofoundation-model-stack:specu-benchmarkfrom
Open
Conversation
Collaborator
|
Thanks @sahilsuneja1 for putting this together! I don't think we need the caller script - it's ultimately just a simple python call with a bunch of arguments, right? I'd just add a comment with a sample call or two (Llama 7b / granite 20b?) to the top of the |
daviswer
reviewed
Mar 25, 2024
| # This example script measures the logical speedup of running a speculator atop a base model. Run as: | ||
| # export CUDA_VISIBLE_DEVICES=1 | ||
| # e.g., #1: torchrun --nproc_per_node=1 benchmark_speculator_logical.py --architecture=paged_llama --variant=7b --model_path=~/models/7B-F --tokenizer=~/models/tokenizer.model --model_source=hf --speculator_path=~/models/speculator_7B_F.pth --compile | ||
| # e.g., #2: torchrun --nproc_per_node=1 benchmark_speculator_logical.py --architecture=paged_gpt_bigcode --variant=ibm.20b --model_path=~/models/granite-20b-instruct --tokenizer=~/models/granite-20b-instruct --model_source=hf --speculator_path=~/models/speculator_granite20B.pth --n_predict=4 --threshes=[6,4,3,3] |
Collaborator
There was a problem hiding this comment.
Needs a --data_path and --subdata!
Collaborator
Author
There was a problem hiding this comment.
Thanks, fixed with fake values
daviswer
approved these changes
Mar 25, 2024
Collaborator
Author
|
Updated formatting |
Collaborator
|
Further CI complaints. I'll try and figure out how to get them running automatically for you rather than having to wait for my explicit go-ahead |
Collaborator
Author
|
Fixed isort issues |
Collaborator
Author
|
Dunno how to fix the mypy errors |
Contributor
|
There are some suggestions here on how to fix some of these, possibly via a fully qualified import? @afrittoli may also have suggestions. pasting the mypy errors here, for easier reference: |
Collaborator
|
Looks like a bunch of that is because this is relying on the paged attention branch, which hasn't fully landed in |
lchu6
changed the title
Collaborator
Contributor
we can either point directly at the github URL within requirements.txt (see some examples here: https://stackoverflow.com/questions/16584552/how-to-state-in-requirements-txt-a-direct-github-source) or we could publish fms-extras to pypi. probably the first option is simpler/easier |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
@daviswer: should we also add the caller script?