You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Run make clean from the root of the pybuda repository.
Run the the code source env_for_wormhole_b0.sh , it compiles pybuda with all necessary submodules for wormhole_b0 arch, one should see activated python_env
When the compilation is finished, go ahead and clone benchmarking repository, and pip install -r requirements.txt , with the same activated python_env.
Next run the falcon model using : python benchmark.py -d tt -m falcon -c 7b-instruct --task hellaswag -df Fp16_b -mb 32 --loop_count 1 --save_output
Expected behavior
The model should without any errors , and produce a benchmarking results similar or close to previous results for the falcon model.
Bug log
backend//cluster_desc.yaml' with timeout 120s Detecting chips (found 1) INFO:benchmark.common.benchmark_run: PyBUDA is not installed as a wheel dist, get git hash from submodule. INFO:__main__: creating benchmarking run: falcon INFO:__main__: kwargs: {'training': False, 'microbatch': 32, 'task': 'hellaswag', 'device': 'tt', 'data_type': 'Fp16_b', 'config': '7b-instruct'} Traceback (most recent call last): File "benchmark.py", line 573, in <module> model, generator, eval_fn = models[args.model]["func"](benchmark_run=benchmark_run, **kwargs) File "/home/aramchandran/Work/benchmarking/benchmark/common/common.py", line 30, in wrapper model, generator, eval_fn = model_func(*args, benchmark_run=benchmark_run, **kwargs) File "/home/aramchandran/Work/benchmarking/benchmark/models/falcon/falcon.py", line 79, in falcon from benchmark.models.falcon.utils.model import Falcon File "/home/aramchandran/Work/benchmarking/benchmark/models/falcon/utils/model.py", line 16, in <module> from transformers.generation.utils import top_k_top_p_filtering ImportError: cannot import name 'top_k_top_p_filtering' from 'transformers.generation.utils' (/home/aramchandran/Work/pybuda/build/python_env/lib/python3.8/site-packages/transformers/generation/utils.py)
System:
Device: Wormhole N150
Host information: [e.g. Intel Xeon]
OS : Ubuntu 20.04]
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered:
Looks like the function 'top_k_top_p_filtering' from 'transformers.generation.utils' method was deprecated from transformers >= 4.39. Current build of pybuda makes use of transformers -= 4.41.0.
Describe the bug
When running the models falcon to benchmark it throws a error.
To Reproduce
Steps to reproduce the behavior:
git submodule sync
andgit submodule update --init --recursive -f
vcanic/nvukobrat_rc_240617
make clean
from the root of the pybuda repository.source env_for_wormhole_b0.sh
, it compiles pybuda with all necessary submodules for wormhole_b0 arch, one should see activated python_envbenchmarking
repository, and pip install -r requirements.txt , with the same activated python_env.python benchmark.py -d tt -m falcon -c 7b-instruct --task hellaswag -df Fp16_b -mb 32 --loop_count 1 --save_output
Expected behavior
The model should without any errors , and produce a benchmarking results similar or close to previous results for the falcon model.
Bug log
System:
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered: