|
checkpoints: null |
|
data: null |
|
experiment_logger: null |
|
general: null |
|
kill_switch_path: null |
|
lighteval: |
|
batch_size: 24 |
|
checkpoints_path: null |
|
generation: null |
|
logging: |
|
hub_repo_details: null |
|
hub_repo_results: null |
|
hub_repo_tensorboard: HuggingFaceBR4/thomwolf-nanotron-mistral-7b |
|
local_output_path: /scratch/thomwolf/lighteval/nanotron-mistral-7b |
|
push_details_to_hub: false |
|
push_results_to_hub: false |
|
push_results_to_tensorboard: true |
|
tensorboard_metric_prefix: e |
|
parallelism: |
|
dp: 4 |
|
pp: 1 |
|
pp_engine: 1f1b |
|
recompute_granularity: null |
|
tp: 2 |
|
tp_linear_async_communication: false |
|
tp_mode: ALL_REDUCE |
|
slurm: null |
|
slurm_script_dir: null |
|
slurm_template: null |
|
tasks: |
|
custom_tasks_file: ./custom_evaluation_tasks.py |
|
dataset_loading_processes: 8 |
|
max_samples: 1000 |
|
multichoice_continuations_start_space: null |
|
no_multichoice_continuations_start_space: null |
|
num_fewshot_seeds: null |
|
tasks: early-signal |
|
logging: null |
|
model: null |
|
optimizer: null |
|
parallelism: null |
|
profiler: null |
|
s3_upload: null |
|
tokenizer: null |
|
tokens: null |
|
|