Failing end2end test test_initialization.py must be fixed #331

COVERAGE_FILE=.coverage_reports/.coverage.part9 CUDA_VISIBLE_DEVICES=$DEV0 coverage run --rcfile=.coveragerc --parallel $(which torchrun) --rdzv-endpoint localhost:29503 --nnodes 1 --nproc_per_node 1 $(which pytest) tests/test_initialization.py $COVERAGE

E       pydantic_core._pydantic_core.ValidationError: 1 validation error for ComposedModelInitializationConfig
E         Value error, hidden_dim must be specified when std is 'auto' [type=value_error, input_value={'model_type': 'gpt2', 'w...rs': 12, 'strict': True}, input_type=dict]
E           For further information visit https://errors.pydantic.dev/2.10/v/value_error

FAILED tests/test_initialization.py::test_nr_parameters_per_initialization_group[gpt2] - AssertionError: nr_parameters for gpt2/weight-normal = 78741504 should be 40108032Actual - Expected: 38633472
FAILED tests/test_initialization.py::test_statistical_distribution_for_each_initialization_group[gpt2-plain-0.02-True] - AssertionError: standard deviation for gpt2/weight-normal = 0.020414261147379875 should be close to 0.019999999552965164
FAILED tests/test_initialization.py::test_statistical_distribution_for_each_initialization_group[gpt2-scaled-0.02-True] - AssertionError: standard deviation for gpt2/weight-normal = 0.020413096994161606 should be close to 0.019999999552965164
FAILED tests/test_initialization.py::test_statistical_distribution_for_each_initialization_group[gpt2-scaled_embed-0.02-True] - AssertionError: standard deviation for gpt2/weight-normal = 0.020415905863046646 should be close to 0.019999999552965164
FAILED tests/test_initialization.py::test_statistical_distribution_for_each_initialization_group[gpt2-plain-auto-True] - pydantic_core._pydantic_core.ValidationError: 1 validation error for ComposedModelInitializationConfig
FAILED tests/test_initialization.py::test_statistical_distribution_for_each_initialization_group[gpt2-scaled-auto-True] - pydantic_core._pydantic_core.ValidationError: 1 validation error for ComposedModelInitializationConfig
FAILED tests/test_initialization.py::test_statistical_distribution_for_each_initialization_group[gpt2-scaled_embed-auto-True] - pydantic_core._pydantic_core.ValidationError: 1 validation error for ComposedModelInitializationConfig