Datasets:

AverageMetaheuristicsEnjoyer
/

mobe-factor-cache-gpt-oss-20b-b16

Tags:

mobe

moe-compression

License:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Preview

Duplicate

Split (1)

train

The full dataset viewer is not available (click to read why). Only showing a preview of the rows.

The dataset generation failed because of a cast error

Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 3 new columns ({'error_message', 'status', 'error_type'}) and 296 missing columns ({'cos_delta_layer_og_vs_method_num_tokens', 'target_layer_compression_ratio', 'cos_after_mha_residual_og_vs_method_median', 'cos_after_mlp_og_vs_method_num_tokens', 'norm_ratio_delta_mlp_to_layer_method_over_og_num_tokens', 'rel_diff_norm_after_mha_residual_og_vs_method_std', 'diff_norm_delta_mha_og_vs_method_mean', 'torch_compile_error_traceback', 'norm_after_mha_residual_method_median', 'cos_after_mlp_og_vs_method_mean', 'norm_delta_mlp_to_layer_og_median', 'ppl_wikitext2', 'overall_no_embeddings_storage_bits', 'angle_deg_after_layer_og_vs_method_mean', 'overall_with_embeddings_storage_bits', 'rel_diff_norm_delta_mlp_to_layer_og_vs_method_median', 'norm_after_layer_method_std', 'target_layer_error_detail', 'rel_diff_norm_after_mha_og_vs_method_std', 'norm_after_layer_og_mean', 'cos_before_layer_og_vs_method_mean', 'norm_delta_layer_method_num_tokens', 'norm_ratio_after_mha_residual_method_over_og_num_tokens', 'diff_norm_delta_layer_og_vs_method_num_tokens', 'rel_diff_norm_after_mha_og_vs_method_median', 'angle_deg_after_mha_og_vs_method_num_tokens', 'norm_after_mlp_method_median', 'rel_diff_norm_before_layer_og_vs_method_std', 'cos_delta_mlp_to_layer_og_vs_method_num_tokens', 'norm_delta_layer_method_median', 'diff_norm_after_mlp_og_vs_method_median', 'norm_ratio_delta_mha_method_over_og_mean', 'target_layer_max_abs_error', 'norm_delta_mha_og_mean', 'rel_diff_norm_delta_layer_og_vs_method_std', 'norm_after_mlp_og_mean', 'diff_norm_after_mha_residual_og_vs_method_num_tokens'
...
s', 'rel_diff_norm_delta_mha_og_vs_method_mean', 'rel_diff_norm_after_mha_residual_og_vs_method_num_tokens', 'norm_delta_mlp_to_layer_og_num_tokens', 'cos_after_mha_residual_og_vs_method_std', 'norm_delta_mlp_to_layer_method_std', 'norm_ratio_after_mlp_method_over_og_std', 'mha_in_affected_blocks_dense_bits', 'diff_norm_before_layer_og_vs_method_median', 'norm_ratio_delta_mlp_to_layer_method_over_og_mean', 'superweight_abs_error_mean', 'rel_diff_norm_after_mlp_og_vs_method_num_tokens', 'mlp_in_affected_blocks_compression_ratio', 'mha_in_affected_blocks_storage_bits', 'norm_ratio_delta_layer_method_over_og_median', 'overall_no_embeddings_compression_ratio', 'norm_ratio_delta_layer_method_over_og_num_tokens', 'angle_deg_delta_mlp_to_layer_og_vs_method_median', 'rel_diff_norm_after_mlp_og_vs_method_mean', 'angle_deg_delta_mha_og_vs_method_std', 'norm_after_mha_residual_og_num_tokens', 'affected_modules_compression_ratio', 'diff_norm_after_mha_og_vs_method_std', 'superweight_abs_error_max', 'angle_deg_after_mha_residual_og_vs_method_std', 'diff_norm_after_layer_og_vs_method_median', 'ppl_requested', 'rel_diff_norm_delta_layer_og_vs_method_median', 'cos_after_mlp_og_vs_method_std', 'norm_ratio_delta_mha_method_over_og_median', 'norm_after_mha_residual_method_num_tokens', 'affected_blocks_storage_bits', 'diff_norm_before_layer_og_vs_method_num_tokens', 'angle_deg_before_layer_og_vs_method_num_tokens', 'angle_deg_delta_mlp_to_layer_og_vs_method_mean', 'dense_reconstruction_records'}).

This happened while the csv dataset builder was generating data using

hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16/openai_gpt-oss-20b__partial_results__status.csv (at revision e802e583f37c3f62b8caadd027e5c7ec956b50d3), ['hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16@e802e583f37c3f62b8caadd027e5c7ec956b50d3/gpt_oss_20b_mobe_b16_results.csv', 'hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16@e802e583f37c3f62b8caadd027e5c7ec956b50d3/openai_gpt-oss-20b__partial_results.csv', 'hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16@e802e583f37c3f62b8caadd027e5c7ec956b50d3/openai_gpt-oss-20b__partial_results__status.csv']

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1837, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              model_name: string
              target_set: string
              method_label: string
              status: string
              error_type: double
              error_message: double
              rank: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1111
              to
              {'target_set': Value('string'), 'model_arch_key': Value('string'), 'model_family': Value('string'), 'model_name': Value('string'), 'method_label': Value('string'), 'rank': Value('float64'), 'target_modules': Value('string'), 'target_layers': Value('string'), 'storage_records': Value('string'), 'affected_modules_dense_bits': Value('int64'), 'affected_modules_storage_bits': Value('int64'), 'affected_modules_compression_ratio': Value('float64'), 'mha_in_affected_blocks_dense_bits': Value('float64'), 'mha_in_affected_blocks_storage_bits': Value('float64'), 'mha_in_affected_blocks_compression_ratio': Value('float64'), 'mlp_in_affected_blocks_dense_bits': Value('float64'), 'mlp_in_affected_blocks_storage_bits': Value('float64'), 'mlp_in_affected_blocks_compression_ratio': Value('float64'), 'affected_blocks_dense_bits': Value('int64'), 'affected_blocks_storage_bits': Value('int64'), 'affected_blocks_compression_ratio': Value('float64'), 'overall_no_embeddings_dense_bits': Value('int64'), 'overall_no_embeddings_storage_bits': Value('int64'), 'overall_no_embeddings_compression_ratio': Value('float64'), 'overall_with_embeddings_dense_bits': Value('int64'), 'overall_with_embeddings_storage_bits': Value('int64'), 'overall_with_embeddings_compression_ratio': Value('float64'), 'target_layer_bits': Value('int64'), 'overall_model_bits': Value('int64'), 'target_layer_compression_ratio': Value('float64'), 'overall_model_compression_ratio': Value('float64'), 'superweight_abs_error_mean': Value(
              ...
              _og_vs_method_median': Value('float64'), 'angle_deg_delta_mlp_to_layer_og_vs_method_std': Value('float64'), 'angle_deg_delta_mlp_to_layer_og_vs_method_num_tokens': Value('float64'), 'norm_delta_mlp_to_layer_og_mean': Value('float64'), 'norm_delta_mlp_to_layer_og_median': Value('float64'), 'norm_delta_mlp_to_layer_og_std': Value('float64'), 'norm_delta_mlp_to_layer_og_num_tokens': Value('float64'), 'norm_delta_mlp_to_layer_method_mean': Value('float64'), 'norm_delta_mlp_to_layer_method_median': Value('float64'), 'norm_delta_mlp_to_layer_method_std': Value('float64'), 'norm_delta_mlp_to_layer_method_num_tokens': Value('float64'), 'norm_ratio_delta_mlp_to_layer_method_over_og_mean': Value('float64'), 'norm_ratio_delta_mlp_to_layer_method_over_og_median': Value('float64'), 'norm_ratio_delta_mlp_to_layer_method_over_og_std': Value('float64'), 'norm_ratio_delta_mlp_to_layer_method_over_og_num_tokens': Value('float64'), 'diff_norm_delta_mlp_to_layer_og_vs_method_mean': Value('float64'), 'diff_norm_delta_mlp_to_layer_og_vs_method_median': Value('float64'), 'diff_norm_delta_mlp_to_layer_og_vs_method_std': Value('float64'), 'diff_norm_delta_mlp_to_layer_og_vs_method_num_tokens': Value('float64'), 'rel_diff_norm_delta_mlp_to_layer_og_vs_method_mean': Value('float64'), 'rel_diff_norm_delta_mlp_to_layer_og_vs_method_median': Value('float64'), 'rel_diff_norm_delta_mlp_to_layer_og_vs_method_std': Value('float64'), 'rel_diff_norm_delta_mlp_to_layer_og_vs_method_num_tokens': Value('float64')}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1683, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1839, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 3 new columns ({'error_message', 'status', 'error_type'}) and 296 missing columns ({'cos_delta_layer_og_vs_method_num_tokens', 'target_layer_compression_ratio', 'cos_after_mha_residual_og_vs_method_median', 'cos_after_mlp_og_vs_method_num_tokens', 'norm_ratio_delta_mlp_to_layer_method_over_og_num_tokens', 'rel_diff_norm_after_mha_residual_og_vs_method_std', 'diff_norm_delta_mha_og_vs_method_mean', 'torch_compile_error_traceback', 'norm_after_mha_residual_method_median', 'cos_after_mlp_og_vs_method_mean', 'norm_delta_mlp_to_layer_og_median', 'ppl_wikitext2', 'overall_no_embeddings_storage_bits', 'angle_deg_after_layer_og_vs_method_mean', 'overall_with_embeddings_storage_bits', 'rel_diff_norm_delta_mlp_to_layer_og_vs_method_median', 'norm_after_layer_method_std', 'target_layer_error_detail', 'rel_diff_norm_after_mha_og_vs_method_std', 'norm_after_layer_og_mean', 'cos_before_layer_og_vs_method_mean', 'norm_delta_layer_method_num_tokens', 'norm_ratio_after_mha_residual_method_over_og_num_tokens', 'diff_norm_delta_layer_og_vs_method_num_tokens', 'rel_diff_norm_after_mha_og_vs_method_median', 'angle_deg_after_mha_og_vs_method_num_tokens', 'norm_after_mlp_method_median', 'rel_diff_norm_before_layer_og_vs_method_std', 'cos_delta_mlp_to_layer_og_vs_method_num_tokens', 'norm_delta_layer_method_median', 'diff_norm_after_mlp_og_vs_method_median', 'norm_ratio_delta_mha_method_over_og_mean', 'target_layer_max_abs_error', 'norm_delta_mha_og_mean', 'rel_diff_norm_delta_layer_og_vs_method_std', 'norm_after_mlp_og_mean', 'diff_norm_after_mha_residual_og_vs_method_num_tokens'
              ...
              s', 'rel_diff_norm_delta_mha_og_vs_method_mean', 'rel_diff_norm_after_mha_residual_og_vs_method_num_tokens', 'norm_delta_mlp_to_layer_og_num_tokens', 'cos_after_mha_residual_og_vs_method_std', 'norm_delta_mlp_to_layer_method_std', 'norm_ratio_after_mlp_method_over_og_std', 'mha_in_affected_blocks_dense_bits', 'diff_norm_before_layer_og_vs_method_median', 'norm_ratio_delta_mlp_to_layer_method_over_og_mean', 'superweight_abs_error_mean', 'rel_diff_norm_after_mlp_og_vs_method_num_tokens', 'mlp_in_affected_blocks_compression_ratio', 'mha_in_affected_blocks_storage_bits', 'norm_ratio_delta_layer_method_over_og_median', 'overall_no_embeddings_compression_ratio', 'norm_ratio_delta_layer_method_over_og_num_tokens', 'angle_deg_delta_mlp_to_layer_og_vs_method_median', 'rel_diff_norm_after_mlp_og_vs_method_mean', 'angle_deg_delta_mha_og_vs_method_std', 'norm_after_mha_residual_og_num_tokens', 'affected_modules_compression_ratio', 'diff_norm_after_mha_og_vs_method_std', 'superweight_abs_error_max', 'angle_deg_after_mha_residual_og_vs_method_std', 'diff_norm_after_layer_og_vs_method_median', 'ppl_requested', 'rel_diff_norm_delta_layer_og_vs_method_median', 'cos_after_mlp_og_vs_method_std', 'norm_ratio_delta_mha_method_over_og_median', 'norm_after_mha_residual_method_num_tokens', 'affected_blocks_storage_bits', 'diff_norm_before_layer_og_vs_method_num_tokens', 'angle_deg_before_layer_og_vs_method_num_tokens', 'angle_deg_delta_mlp_to_layer_og_vs_method_mean', 'dense_reconstruction_records'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16/openai_gpt-oss-20b__partial_results__status.csv (at revision e802e583f37c3f62b8caadd027e5c7ec956b50d3), ['hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16@e802e583f37c3f62b8caadd027e5c7ec956b50d3/gpt_oss_20b_mobe_b16_results.csv', 'hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16@e802e583f37c3f62b8caadd027e5c7ec956b50d3/openai_gpt-oss-20b__partial_results.csv', 'hf://datasets/AverageMetaheuristicsEnjoyer/mobe-factor-cache-gpt-oss-20b-b16@e802e583f37c3f62b8caadd027e5c7ec956b50d3/openai_gpt-oss-20b__partial_results__status.csv']
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

target_set string	model_arch_key string	model_family string	model_name string	method_label string	rank float64	target_modules string	target_layers string	storage_records string	affected_modules_dense_bits int64	affected_modules_storage_bits int64	affected_modules_compression_ratio float64	mha_in_affected_blocks_dense_bits float64	mha_in_affected_blocks_storage_bits float64	mha_in_affected_blocks_compression_ratio null	mlp_in_affected_blocks_dense_bits float64	mlp_in_affected_blocks_storage_bits float64	mlp_in_affected_blocks_compression_ratio null	affected_blocks_dense_bits int64	affected_blocks_storage_bits int64	affected_blocks_compression_ratio float64	overall_no_embeddings_dense_bits int64	overall_no_embeddings_storage_bits int64	overall_no_embeddings_compression_ratio float64	overall_with_embeddings_dense_bits int64	overall_with_embeddings_storage_bits int64	overall_with_embeddings_compression_ratio float64	target_layer_bits int64	overall_model_bits int64	target_layer_compression_ratio float64	overall_model_compression_ratio float64	superweight_abs_error_mean null	superweight_abs_error_max null	superweight_rel_error_mean null	superweight_rel_error_max null	superweight_error_detail string	target_layer_fro_rel_error float64	target_layer_max_abs_error float64	target_layer_error_detail string	mean_top_outlier_rel_error float64	mean_random_rel_error float64	decomposition_apply_time_s float64	quantization_apply_time_s float64	baseline_benchmarks_cached_per_model bool	ppl_requested bool	lm_eval_requested bool	hotpot_requested bool	generation_examples_requested bool	torch_compile_requested bool	torch_compile_applied bool	torch_compile_backend null	torch_compile_mode null	torch_compile_error null	torch_compile_error_traceback null	torch_compile_skip_reason null	torch_compile_device_summary string	ppl_wikitext2 float64	ppl_c4 float64	wikitext2_ppl float64	lm_eval_attempted bool	lm_eval_tasks_configured string	lm_eval_skipped_reason string	generation_examples string	generation_preview_prompt string	generation_preview_text string	resume_key string	moe_experts_in_affected_blocks_dense_bits float64	moe_experts_in_affected_blocks_storage_bits float64	moe_experts_in_affected_blocks_compression_ratio float64	recipe string	plan_stats string	benchmarked_with_dense_reconstruction bool	dense_reconstruction_records string	activation_geometry_stats string	activation_geometry_per_layer_stats string	cos_before_layer_og_vs_method_mean float64	cos_before_layer_og_vs_method_median float64	cos_before_layer_og_vs_method_std float64	cos_before_layer_og_vs_method_num_tokens float64	angle_deg_before_layer_og_vs_method_mean float64	angle_deg_before_layer_og_vs_method_median float64	angle_deg_before_layer_og_vs_method_std float64	angle_deg_before_layer_og_vs_method_num_tokens float64	norm_before_layer_og_mean float64	norm_before_layer_og_median float64	norm_before_layer_og_std float64	norm_before_layer_og_num_tokens float64	norm_before_layer_method_mean float64	norm_before_layer_method_median float64	norm_before_layer_method_std float64	norm_before_layer_method_num_tokens float64	norm_ratio_before_layer_method_over_og_mean float64	norm_ratio_before_layer_method_over_og_median float64	norm_ratio_before_layer_method_over_og_std float64	norm_ratio_before_layer_method_over_og_num_tokens float64	diff_norm_before_layer_og_vs_method_mean float64	diff_norm_before_layer_og_vs_method_median float64	diff_norm_before_layer_og_vs_method_std float64	diff_norm_before_layer_og_vs_method_num_tokens float64	rel_diff_norm_before_layer_og_vs_method_mean float64	rel_diff_norm_before_layer_og_vs_method_median float64	rel_diff_norm_before_layer_og_vs_method_std float64	rel_diff_norm_before_layer_og_vs_method_num_tokens float64	cos_after_mha_og_vs_method_mean float64	cos_after_mha_og_vs_method_median float64	cos_after_mha_og_vs_method_std float64	cos_after_mha_og_vs_method_num_tokens float64	angle_deg_after_mha_og_vs_method_mean float64	angle_deg_after_mha_og_vs_method_median float64	angle_deg_after_mha_og_vs_method_std float64	angle_deg_after_mha_og_vs_method_num_tokens float64	norm_after_mha_og_mean float64	norm_after_mha_og_median float64	norm_after_mha_og_std float64	norm_after_mha_og_num_tokens float64	norm_after_mha_method_mean float64	norm_after_mha_method_median float64	norm_after_mha_method_std float64	norm_after_mha_method_num_tokens float64	norm_ratio_after_mha_method_over_og_mean float64	norm_ratio_after_mha_method_over_og_median float64	norm_ratio_after_mha_method_over_og_std float64	norm_ratio_after_mha_method_over_og_num_tokens float64	diff_norm_after_mha_og_vs_method_mean float64	diff_norm_after_mha_og_vs_method_median float64	diff_norm_after_mha_og_vs_method_std float64	diff_norm_after_mha_og_vs_method_num_tokens float64	rel_diff_norm_after_mha_og_vs_method_mean float64	rel_diff_norm_after_mha_og_vs_method_median float64	rel_diff_norm_after_mha_og_vs_method_std float64	rel_diff_norm_after_mha_og_vs_method_num_tokens float64	cos_after_mha_residual_og_vs_method_mean float64	cos_after_mha_residual_og_vs_method_median float64	cos_after_mha_residual_og_vs_method_std float64	cos_after_mha_residual_og_vs_method_num_tokens float64	angle_deg_after_mha_residual_og_vs_method_mean float64	angle_deg_after_mha_residual_og_vs_method_median float64	angle_deg_after_mha_residual_og_vs_method_std float64	angle_deg_after_mha_residual_og_vs_method_num_tokens float64	norm_after_mha_residual_og_mean float64	norm_after_mha_residual_og_median float64	norm_after_mha_residual_og_std float64	norm_after_mha_residual_og_num_tokens float64	norm_after_mha_residual_method_mean float64	norm_after_mha_residual_method_median float64	norm_after_mha_residual_method_std float64	norm_after_mha_residual_method_num_tokens float64	norm_ratio_after_mha_residual_method_over_og_mean float64	norm_ratio_after_mha_residual_method_over_og_median float64	norm_ratio_after_mha_residual_method_over_og_std float64	norm_ratio_after_mha_residual_method_over_og_num_tokens float64	diff_norm_after_mha_residual_og_vs_method_mean float64	diff_norm_after_mha_residual_og_vs_method_median float64	diff_norm_after_mha_residual_og_vs_method_std float64	diff_norm_after_mha_residual_og_vs_method_num_tokens float64	rel_diff_norm_after_mha_residual_og_vs_method_mean float64	rel_diff_norm_after_mha_residual_og_vs_method_median float64	rel_diff_norm_after_mha_residual_og_vs_method_std float64	rel_diff_norm_after_mha_residual_og_vs_method_num_tokens float64	cos_after_mlp_og_vs_method_mean float64	cos_after_mlp_og_vs_method_median float64	cos_after_mlp_og_vs_method_std float64	cos_after_mlp_og_vs_method_num_tokens float64	angle_deg_after_mlp_og_vs_method_mean float64	angle_deg_after_mlp_og_vs_method_median float64	angle_deg_after_mlp_og_vs_method_std float64	angle_deg_after_mlp_og_vs_method_num_tokens float64	norm_after_mlp_og_mean float64	norm_after_mlp_og_median float64	norm_after_mlp_og_std float64	norm_after_mlp_og_num_tokens float64	norm_after_mlp_method_mean float64	norm_after_mlp_method_median float64	norm_after_mlp_method_std float64	norm_after_mlp_method_num_tokens float64	norm_ratio_after_mlp_method_over_og_mean float64	norm_ratio_after_mlp_method_over_og_median float64	norm_ratio_after_mlp_method_over_og_std float64	norm_ratio_after_mlp_method_over_og_num_tokens float64	diff_norm_after_mlp_og_vs_method_mean float64	diff_norm_after_mlp_og_vs_method_median float64	diff_norm_after_mlp_og_vs_method_std float64	diff_norm_after_mlp_og_vs_method_num_tokens float64	rel_diff_norm_after_mlp_og_vs_method_mean float64	rel_diff_norm_after_mlp_og_vs_method_median float64	rel_diff_norm_after_mlp_og_vs_method_std float64	rel_diff_norm_after_mlp_og_vs_method_num_tokens float64	cos_after_layer_og_vs_method_mean float64	cos_after_layer_og_vs_method_median float64	cos_after_layer_og_vs_method_std float64	cos_after_layer_og_vs_method_num_tokens float64	angle_deg_after_layer_og_vs_method_mean float64	angle_deg_after_layer_og_vs_method_median float64	angle_deg_after_layer_og_vs_method_std float64	angle_deg_after_layer_og_vs_method_num_tokens float64	norm_after_layer_og_mean float64	norm_after_layer_og_median float64	norm_after_layer_og_std float64	norm_after_layer_og_num_tokens float64	norm_after_layer_method_mean float64	norm_after_layer_method_median float64	norm_after_layer_method_std float64	norm_after_layer_method_num_tokens float64	norm_ratio_after_layer_method_over_og_mean float64	norm_ratio_after_layer_method_over_og_median float64	norm_ratio_after_layer_method_over_og_std float64	norm_ratio_after_layer_method_over_og_num_tokens float64	diff_norm_after_layer_og_vs_method_mean float64	diff_norm_after_layer_og_vs_method_median float64	diff_norm_after_layer_og_vs_method_std float64	diff_norm_after_layer_og_vs_method_num_tokens float64	rel_diff_norm_after_layer_og_vs_method_mean float64	rel_diff_norm_after_layer_og_vs_method_median float64	rel_diff_norm_after_layer_og_vs_method_std float64	rel_diff_norm_after_layer_og_vs_method_num_tokens float64	cos_delta_layer_og_vs_method_mean float64	cos_delta_layer_og_vs_method_median float64	cos_delta_layer_og_vs_method_std float64	cos_delta_layer_og_vs_method_num_tokens float64	angle_deg_delta_layer_og_vs_method_mean float64	angle_deg_delta_layer_og_vs_method_median float64	angle_deg_delta_layer_og_vs_method_std float64	angle_deg_delta_layer_og_vs_method_num_tokens float64	norm_delta_layer_og_mean float64	norm_delta_layer_og_median float64	norm_delta_layer_og_std float64	norm_delta_layer_og_num_tokens float64	norm_delta_layer_method_mean float64	norm_delta_layer_method_median float64	norm_delta_layer_method_std float64	norm_delta_layer_method_num_tokens float64	norm_ratio_delta_layer_method_over_og_mean float64	norm_ratio_delta_layer_method_over_og_median float64	norm_ratio_delta_layer_method_over_og_std float64	norm_ratio_delta_layer_method_over_og_num_tokens float64	diff_norm_delta_layer_og_vs_method_mean float64	diff_norm_delta_layer_og_vs_method_median float64	diff_norm_delta_layer_og_vs_method_std float64	diff_norm_delta_layer_og_vs_method_num_tokens float64	rel_diff_norm_delta_layer_og_vs_method_mean float64	rel_diff_norm_delta_layer_og_vs_method_median float64	rel_diff_norm_delta_layer_og_vs_method_std float64	rel_diff_norm_delta_layer_og_vs_method_num_tokens float64	cos_delta_mha_og_vs_method_mean float64	cos_delta_mha_og_vs_method_median float64	cos_delta_mha_og_vs_method_std float64	cos_delta_mha_og_vs_method_num_tokens float64	angle_deg_delta_mha_og_vs_method_mean float64	angle_deg_delta_mha_og_vs_method_median float64	angle_deg_delta_mha_og_vs_method_std float64	angle_deg_delta_mha_og_vs_method_num_tokens float64	norm_delta_mha_og_mean float64	norm_delta_mha_og_median float64	norm_delta_mha_og_std float64	norm_delta_mha_og_num_tokens float64	norm_delta_mha_method_mean float64	norm_delta_mha_method_median float64	norm_delta_mha_method_std float64	norm_delta_mha_method_num_tokens float64	norm_ratio_delta_mha_method_over_og_mean float64	norm_ratio_delta_mha_method_over_og_median float64	norm_ratio_delta_mha_method_over_og_std float64	norm_ratio_delta_mha_method_over_og_num_tokens float64	diff_norm_delta_mha_og_vs_method_mean float64	diff_norm_delta_mha_og_vs_method_median float64	diff_norm_delta_mha_og_vs_method_std float64	diff_norm_delta_mha_og_vs_method_num_tokens float64	rel_diff_norm_delta_mha_og_vs_method_mean float64	rel_diff_norm_delta_mha_og_vs_method_median float64	rel_diff_norm_delta_mha_og_vs_method_std float64	rel_diff_norm_delta_mha_og_vs_method_num_tokens float64	cos_delta_mlp_to_layer_og_vs_method_mean float64	cos_delta_mlp_to_layer_og_vs_method_median float64	cos_delta_mlp_to_layer_og_vs_method_std float64	cos_delta_mlp_to_layer_og_vs_method_num_tokens float64	angle_deg_delta_mlp_to_layer_og_vs_method_mean float64	angle_deg_delta_mlp_to_layer_og_vs_method_median float64	angle_deg_delta_mlp_to_layer_og_vs_method_std float64	angle_deg_delta_mlp_to_layer_og_vs_method_num_tokens float64	norm_delta_mlp_to_layer_og_mean float64	norm_delta_mlp_to_layer_og_median float64	norm_delta_mlp_to_layer_og_std float64	norm_delta_mlp_to_layer_og_num_tokens float64	norm_delta_mlp_to_layer_method_mean float64	norm_delta_mlp_to_layer_method_median float64	norm_delta_mlp_to_layer_method_std float64	norm_delta_mlp_to_layer_method_num_tokens float64	norm_ratio_delta_mlp_to_layer_method_over_og_mean float64	norm_ratio_delta_mlp_to_layer_method_over_og_median float64	norm_ratio_delta_mlp_to_layer_method_over_og_std float64	norm_ratio_delta_mlp_to_layer_method_over_og_num_tokens float64	diff_norm_delta_mlp_to_layer_og_vs_method_mean float64	diff_norm_delta_mlp_to_layer_og_vs_method_median float64	diff_norm_delta_mlp_to_layer_og_vs_method_std float64	diff_norm_delta_mlp_to_layer_og_vs_method_num_tokens float64	rel_diff_norm_delta_mlp_to_layer_og_vs_method_mean float64	rel_diff_norm_delta_mlp_to_layer_og_vs_method_median float64	rel_diff_norm_delta_mlp_to_layer_og_vs_method_std float64	rel_diff_norm_delta_mlp_to_layer_og_vs_method_num_tokens float64
__model_baseline__	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	baseline	null	[]	[]	[]	0	0	null	0	0	null	0	0	null	0	0	null	325,369,979,904	325,369,979,904	1	334,636,114,944	334,636,114,944	1	0	334,636,114,944	null	1	null	null	null	null	[]	null	null	[]	null	null	0	0	false	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	503.871613	269.941589	503.871613	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	['openai/gpt-oss-20b', '__model_baseline__', 'baseline']	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null	null
gpt_oss_20b_mobe_middle_out_layers_12_12	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	['model.layers.12.mlp.experts']	[12]	"[{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	8,493,465,600	6,794,788,864	1.249997	null	null	null	null	null	null	13,170,991,616	11,472,314,880	1.148067	325,369,979,904	323,671,303,168	1.005248	334,636,114,944	332,937,438,208	1.005102	6,794,788,864	332,937,438,208	1.249997	1.005102	null	null	null	null	[]	0.217808	0.130859	"[{'module': 'model.layers.12.mlp.experts', 'fro_rel_error': 0.2178077225031497, 'max_abs_error': 0.(...TRUNCATED)	0.016215	0.311191	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	471.656036	246.069794	471.656036	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_12_12', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	8,493,465,600	6,794,788,864	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.12.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, 'std':(...TRUNCATED)	"{12: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	1.000001	1.000001	0	102,400	0.000007	0	0.000517	102,400	1,344.241577	1,305.57666	1,248.390991	102,400	1,344.241577	1,305.57666	1,248.390991	102,400	1	1	0	102,400	0	0	0	102,400	0	0	0	102,400	1	1	0	102,400	0.00111	0	0.006082	102,400	300.829041	288.386688	77.629799	102,400	300.829041	288.386688	77.629799	102,400	1	1	0	102,400	0	0	0	102,400	0	0	0	102,400	1.000001	1.000001	0	102,400	0.000002	0	0.000321	102,400	1,277.930786	1,249.170166	1,250.1604	102,400	1,277.930786	1,249.170166	1,250.1604	102,400	1	1	0	102,400	0	0	0	102,400	0	0	0	102,400	0.906975	0.926692	0.069975	102,400	23.631529	22.075184	8.255015	102,400	560.145996	544.13092	190.614685	102,400	378.215149	358.430542	117.296257	102,400	0.691647	0.685962	0.120352	102,400	258.550873	250.70163	98.826271	102,400	0.470282	0.457926	0.113403	102,400	0.990503	0.991592	0.005102	102,400	7.645831	7.435006	2.006101	102,400	1,627.13916	1,588.779297	1,248.423096	102,400	1,467.950684	1,429.704712	1,247.200073	102,400	0.902165	0.904971	0.03848	102,400	258.586456	250.745819	98.816788	102,400	0.160336	0.156424	0.046253	102,400	0.840926	0.866436	0.102694	102,400	31.327934	29.952868	10.283725	102,400	490.230713	472.366577	132.352768	102,400	393.960388	385.100098	72.961143	102,400	0.822462	0.822429	0.128651	102,400	258.586456	250.745819	98.816788	102,400	0.528641	0.517869	0.14683	102,400	1	1	0	102,400	0.000009	0	0.000463	102,400	300.837708	288.368835	77.630577	102,400	300.837708	288.368835	77.630577	102,400	1	1	0	102,400	0	0	0	102,400	0	0	0	102,400	0.906932	0.926647	0.069988	102,400	23.637728	22.08197	8.255298	102,400	560.151917	544.191772	190.606232	102,400	378.21579	358.462952	117.29142	102,400	0.691639	0.685911	0.120351	102,400	258.586456	250.745819	98.816788	102,400	0.470355	0.458099	0.113393	102,400
gpt_oss_20b_mobe_middle_out_layers_12_13	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	['model.layers.12.mlp.experts', 'model.layers.13.mlp.experts']	[12, 13]	"[{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	16,986,931,200	13,589,577,728	1.249997	null	null	null	null	null	null	26,341,983,232	22,944,629,760	1.148067	325,369,979,904	321,972,626,432	1.010552	334,636,114,944	331,238,761,472	1.010257	13,589,577,728	331,238,761,472	1.249997	1.010257	null	null	null	null	[]	0.21591	0.130859	"[{'module': 'model.layers.12.mlp.experts', 'fro_rel_error': 0.2178077225031497, 'max_abs_error': 0.(...TRUNCATED)	0.009694	0.299989	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	436.210541	232.704193	436.210541	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_12_13', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	16,986,931,200	13,589,577,728	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.12.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.9952517151832581, 'median': 1.0, 'std': 0.005964030046(...TRUNCATED)	"{12: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	0.995252	1	0.005964	204,800	3.822919	0	4.077614	204,800	1,485.690308	1,421.543945	1,256.391846	204,800	1,406.096069	1,364.161255	1,249.324707	204,800	0.951082	1	0.055976	204,800	129.293228	21.751879	146.966568	204,800	0.080168	0.000522	0.086583	204,800	0.848889	0.999995	0.21971	204,800	21.684525	0.158956	24.872847	204,800	349.424194	332.342621	110.408577	204,800	346.302765	343.936157	90.012657	204,800	1.02193	1	0.230224	204,800	148.953903	0.547475	165.097305	204,800	0.400305	0.002234	0.474036	204,800	0.996656	1	0.004271	204,800	3.239426	0	3.390086	204,800	1,397.004639	1,351.952759	1,252.901978	204,800	1,413.382813	1,363.317505	1,254.150269	204,800	1.011384	1	0.026712	204,800	88.29966	21.545275	95.090271	204,800	0.060025	0.000519	0.063792	204,800	0.916377	0.937304	0.070804	204,800	22.242195	20.396458	8.278032	204,800	596.307007	587.995972	172.380447	204,800	446.304382	424.965515	148.918091	204,800	0.749804	0.755621	0.125968	204,800	245.147156	234.208466	82.325043	204,800	0.426025	0.411062	0.126407	204,800	0.990598	0.991478	0.004622	204,800	7.661973	7.485298	1.773166	204,800	1,766.085938	1,734.776611	1,249.052246	204,800	1,646.48999	1,608.84436	1,253.987427	204,800	0.929948	0.936654	0.042828	204,800	259.099091	251.806488	80.311096	204,800	0.149749	0.143139	0.03971	204,800	0.880171	0.911472	0.0897	204,800	26.890409	24.290432	9.497786	204,800	533.329712	521.180115	129.167236	204,800	516.583679	503.920837	149.452011	204,800	0.972214	0.986274	0.197939	204,800	253.199203	240.639679	84.025581	204,800	0.485934	0.458247	0.150438	204,800	0.848873	0.999972	0.21972	204,800	21.685583	0.318449	24.874226	204,800	349.433136	332.331299	110.409798	204,800	346.31073	343.943115	90.013023	204,800	1.021928	1	0.230224	204,800	148.96608	1.288562	165.106293	204,800	0.400329	0.005265	0.474053	204,800	0.91634	0.937272	0.070816	204,800	22.247841	20.401621	8.278292	204,800	596.310425	588.029907	172.375443	204,800	446.3078	424.963989	148.915695	204,800	0.749808	0.755526	0.12597	204,800	245.185333	234.293274	82.318527	204,800	0.426092	0.411126	0.1264	204,800
gpt_oss_20b_mobe_middle_out_layers_12_14	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	['model.layers.12.mlp.experts', 'model.layers.13.mlp.experts', 'model.layers.14.mlp.experts']	[12, 13, 14]	"[{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	25,480,396,800	20,384,366,592	1.249997	null	null	null	null	null	null	39,512,974,848	34,416,944,640	1.148067	325,369,979,904	320,273,949,696	1.015911	334,636,114,944	329,540,084,736	1.015464	20,384,366,592	329,540,084,736	1.249997	1.015464	null	null	null	null	[]	0.214486	0.146484	"[{'module': 'model.layers.12.mlp.experts', 'fro_rel_error': 0.2178077225031497, 'max_abs_error': 0.(...TRUNCATED)	0.009087	0.298921	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	505.448547	241.988846	505.448547	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_12_14', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	25,480,396,800	20,384,366,592	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.12.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.9937322735786438, 'median': 0.9936278462409973, 'std':(...TRUNCATED)	"{12: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	0.993732	0.993628	0.005822	307,200	5.107984	6.4716	3.891251	307,200	1,625.471191	1,581.55835	1,264.563599	307,200	1,545.740601	1,458.278076	1,260.203125	307,200	0.953299	0.960914	0.048097	307,200	172.732727	211.930298	138.629944	307,200	0.099833	0.122684	0.077682	307,200	0.871622	0.93158	0.184157	307,200	22.048452	21.317474	20.607073	307,200	377.537384	374.607361	104.841316	307,200	384.067505	385.434143	98.380775	307,200	1.03582	1	0.196676	307,200	159.828629	170.787445	138.72966	307,200	0.407089	0.394134	0.392511	307,200	0.994958	0.99459	0.004755	307,200	4.597062	5.96266	3.467484	307,200	1,572.967407	1,483.82373	1,270.757446	307,200	1,588.133789	1,512.848633	1,271.235962	307,200	1.009859	1	0.02415	307,200	139.699707	168.136536	109.146866	307,200	0.08307	0.107288	0.063115	307,200	0.922271	0.938566	0.0613	307,200	21.625963	20.187912	7.358484	307,200	691.107666	667.302734	221.688232	307,200	557.246643	528.021057	217.694199	307,200	0.796914	0.817882	0.141316	307,200	266.833069	261.246704	83.936775	307,200	0.404573	0.383973	0.128738	307,200	0.989717	0.990492	0.004721	307,200	8.030582	7.907134	1.778456	307,200	2,003.269165	1,880.019043	1,287.41272	307,200	1,884.954346	1,788.7146	1,290.898193	307,200	0.937745	0.945501	0.039008	307,200	300.859467	290.664032	98.329865	307,200	0.152393	0.148171	0.036323	307,200	0.904182	0.933792	0.081619	307,200	23.745049	20.966202	9.164392	307,200	664.235535	596.437683	227.035263	307,200	654.782288	633.835693	242.464401	307,200	0.984849	0.999227	0.17555	307,200	263.787323	256.772156	78.108833	307,200	0.428376	0.387617	0.161478	307,200	0.871593	0.931564	0.184163	307,200	22.052776	21.320053	20.607307	307,200	377.546814	374.596191	104.843163	307,200	384.075378	385.432007	98.380653	307,200	1.035817	1	0.196677	307,200	159.857681	170.866089	138.738663	307,200	0.407166	0.394173	0.39251	307,200	0.922238	0.938549	0.061312	307,200	21.631023	20.190674	7.358778	307,200	691.110779	667.349243	221.687134	307,200	557.250854	527.991333	217.693451	307,200	0.796919	0.817918	0.141345	307,200	266.876129	261.260254	83.937515	307,200	0.404639	0.384101	0.12876	307,200
gpt_oss_20b_mobe_middle_out_layers_12_15	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	"['model.layers.12.mlp.experts', 'model.layers.13.mlp.experts', 'model.layers.14.mlp.experts', 'mode(...TRUNCATED)	[12, 13, 14, 15]	"[{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	33,973,862,400	27,179,155,456	1.249997	null	null	null	null	null	null	52,683,966,464	45,889,259,520	1.148067	325,369,979,904	318,575,272,960	1.021328	334,636,114,944	327,841,408,000	1.020726	27,179,155,456	327,841,408,000	1.249997	1.020726	null	null	null	null	[]	0.213755	0.146484	"[{'module': 'model.layers.12.mlp.experts', 'fro_rel_error': 0.2178077225031497, 'max_abs_error': 0.(...TRUNCATED)	0.007014	0.307724	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	482.517914	236.946518	482.517914	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_12_15', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	33,973,862,400	27,179,155,456	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.12.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.9922881126403809, 'median': 0.9920768737792969, 'std':(...TRUNCATED)	"{12: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	0.992288	0.992077	0.006045	409,600	6.022939	7.217258	3.803171	409,600	1,838.512329	1,734.224609	1,309.245728	409,600	1,749.775879	1,611.044922	1,301.633179	409,600	0.953309	0.957632	0.043219	409,600	225.644623	251.11673	155.638809	409,600	0.114294	0.135262	0.073102	409,600	0.864282	0.91054	0.166129	409,600	24.389622	24.419916	18.890156	409,600	424.528442	415.595032	146.230972	409,600	454.396332	428.678345	163.947266	409,600	1.076783	1.015732	0.21085	409,600	208.005432	212.352173	156.066116	409,600	0.46728	0.460242	0.37312	409,600	0.992599	0.993182	0.006465	409,600	5.849183	6.694598	3.806266	409,600	1,794.092651	1,657.943848	1,317.379883	409,600	1,824.80896	1,683.112305	1,327.060181	409,600	1.015169	1.008147	0.026697	409,600	210.090332	204.556595	158.536469	409,600	0.106006	0.120426	0.069433	409,600	0.910974	0.929274	0.065768	409,600	23.175055	21.678028	7.857037	409,600	758.87439	743.864929	243.536896	409,600	615.716187	611.9646	232.976822	409,600	0.803166	0.81702	0.130987	409,600	312.617706	293.290985	117.679428	409,600	0.422794	0.406099	0.127796	409,600	0.987645	0.98905	0.0065	409,600	8.742085	8.486712	2.215853	409,600	2,260.985107	2,082.342285	1,346.817871	409,600	2,140.035156	1,986.073486	1,348.24646	409,600	0.942631	0.949068	0.03666	409,600	368.552429	336.767212	152.397125	409,600	0.162173	0.157902	0.039621	409,600	0.909434	0.932017	0.072524	409,600	23.303085	21.248486	8.209533	409,600	766.412109	739.088623	277.237915	409,600	761.082031	761.718567	290.557953	409,600	0.99095	1.001202	0.155649	409,600	301.108398	283.159424	102.186798	409,600	0.418232	0.385291	0.145178	409,600	0.864249	0.910462	0.166137	409,600	24.39414	24.430706	18.890491	409,600	424.537994	415.620605	146.231781	409,600	454.404755	428.67157	163.948105	409,600	1.07678	1.015689	0.210851	409,600	208.040237	212.390564	156.078613	409,600	0.467359	0.4603	0.373122	409,600	0.91094	0.929223	0.06578	409,600	23.179922	21.685925	7.857218	409,600	758.875305	743.829102	243.53447	409,600	615.720642	611.996704	232.976257	409,600	0.803173	0.817023	0.131011	409,600	312.662384	293.350433	117.681602	409,600	0.422857	0.40623	0.127813	409,600
gpt_oss_20b_mobe_middle_out_layers_12_16	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	"['model.layers.12.mlp.experts', 'model.layers.13.mlp.experts', 'model.layers.14.mlp.experts', 'mode(...TRUNCATED)	[12, 13, 14, 15, 16]	"[{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	42,467,328,000	33,973,944,320	1.249997	null	null	null	null	null	null	65,854,958,080	57,361,574,400	1.148067	325,369,979,904	316,876,596,224	1.026803	334,636,114,944	326,142,731,264	1.026042	33,973,944,320	326,142,731,264	1.249997	1.026042	null	null	null	null	[]	0.214856	0.146484	"[{'module': 'model.layers.12.mlp.experts', 'fro_rel_error': 0.2178077225031497, 'max_abs_error': 0.(...TRUNCATED)	0.005863	0.311513	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	567.910706	268.561157	567.910706	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_12_16', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	42,467,328,000	33,973,944,320	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.12.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.9901159405708313, 'median': 0.9906193017959595, 'std':(...TRUNCATED)	"{12: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	0.990116	0.990619	0.007631	512,000	6.99367	7.85408	4.019434	512,000	2,077.636475	1,880.317871	1,377.423218	512,000	1,980.876465	1,789.062012	1,366.467773	512,000	0.954105	0.957289	0.040022	512,000	294.841949	290.575775	200.78067	512,000	0.129738	0.146434	0.073918	512,000	0.873582	0.913625	0.15093	512,000	24.266645	23.988852	17.083473	512,000	484.837067	454.861633	185.238358	512,000	533.61261	475.628845	228.124603	512,000	1.09553	1.059365	0.19979	512,000	236.792908	247.134888	156.431793	512,000	0.471564	0.466656	0.338235	512,000	0.991294	0.991798	0.006807	512,000	6.562831	7.343447	3.771725	512,000	2,085.872314	1,889.704102	1,423.619629	512,000	2,123.620605	1,907.816895	1,436.973999	512,000	1.01622	1.011469	0.025894	512,000	275.734924	241.849823	196.953552	512,000	0.118483	0.132054	0.068348	512,000	0.909915	0.926809	0.06371	512,000	23.376507	22.057207	7.691873	512,000	875.108215	832.12854	344.380554	512,000	725.846863	697.406006	320.776794	512,000	0.817859	0.833782	0.126203	512,000	359.995728	329.306519	154.158646	512,000	0.422524	0.406956	0.123832	512,000	0.986559	0.987978	0.007004	512,000	9.118278	8.893112	2.315312	512,000	2,639.976074	2,442.887695	1,526.758179	512,000	2,511.206543	2,325.142578	1,520.254395	512,000	0.94642	0.952282	0.036015	512,000	448.52478	389.660309	220.488434	512,000	0.167236	0.163578	0.040293	512,000	0.918555	0.938992	0.068181	512,000	21.998846	20.117102	8.000486	512,000	941.85791	895.368652	444.132507	512,000	940.47052	902.00061	455.96228	512,000	0.995018	1.004526	0.142882	512,000	338.420837	314.5979	128.71347	512,000	0.394822	0.363429	0.141726	512,000	0.873544	0.913585	0.150937	512,000	24.271914	23.994474	17.083664	512,000	484.847748	454.86319	185.240005	512,000	533.621277	475.627441	228.125168	512,000	1.095526	1.059391	0.19979	512,000	236.839951	247.206451	156.45079	512,000	0.471653	0.466775	0.338237	512,000	0.909881	0.926797	0.063722	512,000	23.381392	22.059034	7.692056	512,000	875.111328	832.12439	344.383575	512,000	725.853333	697.416748	320.780365	512,000	0.817865	0.833792	0.126225	512,000	360.05249	329.352844	154.180725	512,000	0.42259	0.407039	0.123847	512,000
gpt_oss_20b_mobe_middle_out_layers_12_17	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	"['model.layers.12.mlp.experts', 'model.layers.13.mlp.experts', 'model.layers.14.mlp.experts', 'mode(...TRUNCATED)	[12, 13, 14, 15, 16, 17]	"[{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	50,960,793,600	40,768,733,184	1.249997	null	null	null	null	null	null	79,025,949,696	68,833,889,280	1.148067	325,369,979,904	315,177,919,488	1.032337	334,636,114,944	324,444,054,528	1.031414	40,768,733,184	324,444,054,528	1.249997	1.031414	null	null	null	null	[]	0.214126	0.146484	"[{'module': 'model.layers.12.mlp.experts', 'fro_rel_error': 0.2178077225031497, 'max_abs_error': 0.(...TRUNCATED)	0.004886	0.308545	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	554.878906	268.462799	554.878906	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_12_17', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	50,960,793,600	40,768,733,184	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.12.mlp.experts', 'layer_idx': 12, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.12.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.9887990355491638, 'median': 0.9894231557846069, 'std':(...TRUNCATED)	"{12: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	0.988799	0.989423	0.008122	614,400	7.598566	8.340637	4.001858	614,400	2,424.020508	2,082.993652	1,560.583618	614,400	2,316.712402	1,986.773926	1,541.058716	614,400	0.95535	0.958758	0.038466	614,400	373.77063	336.767212	261.636292	614,400	0.139364	0.154164	0.07237	614,400	0.860153	0.901396	0.150699	614,400	26.217691	25.657772	16.889408	614,400	560.583862	498.597229	250.892471	614,400	631.70166	541.079041	318.97818	614,400	1.114725	1.078962	0.212993	614,400	309.229462	284.423157	230.373734	614,400	0.516193	0.50281	0.341219	614,400	0.989936	0.990498	0.00756	614,400	7.189078	7.904626	3.816613	614,400	2,428.504639	2,069.985352	1,588.734375	614,400	2,505.704102	2,086.943848	1,646.026733	614,400	1.024491	1.017198	0.033652	614,400	365.851654	315.526764	277.676086	614,400	0.132052	0.143147	0.071432	614,400	0.911342	0.92737	0.0616	614,400	23.214365	21.971487	7.530819	614,400	1,025.315186	919.604309	685.414978	614,400	841.243835	776.705322	610.889221	614,400	0.813372	0.824404	0.119735	614,400	415.615906	372.454468	201.360504	614,400	0.419925	0.405191	0.120083	614,400	0.986025	0.987438	0.007585	614,400	9.301551	9.091115	2.348477	614,400	3,080.009277	2,687.3125	1,716.786865	614,400	2,955.820801	2,560.935547	1,718.579956	614,400	0.95215	0.956773	0.037343	614,400	530.103577	465.152985	282.469788	614,400	0.169253	0.165553	0.041078	614,400	0.923029	0.941531	0.064127	614,400	21.412165	19.689676	7.655539	614,400	1,124.123291	992.535767	761.697144	614,400	1,118.919434	998.41687	735.0448	614,400	0.994465	1.0007	0.132957	614,400	389.156982	353.327087	178.619888	614,400	0.383218	0.353777	0.136017	614,400	0.860115	0.901338	0.150705	614,400	26.222685	25.665485	16.889332	614,400	560.596741	498.629333	250.897232	614,400	631.711487	541.108398	318.981049	614,400	1.114719	1.078974	0.212992	614,400	309.28006	284.516602	230.390564	614,400	0.516276	0.502916	0.341215	614,400	0.911306	0.927348	0.061612	614,400	23.219505	21.974823	7.531025	614,400	1,025.319824	919.672241	685.41803	614,400	841.250183	776.772339	610.890503	614,400	0.813377	0.824384	0.119756	614,400	415.686676	372.523743	201.394867	614,400	0.419993	0.405244	0.120096	614,400
gpt_oss_20b_mobe_middle_out_layers_11_17	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	"['model.layers.11.mlp.experts', 'model.layers.12.mlp.experts', 'model.layers.13.mlp.experts', 'mode(...TRUNCATED)	[11, 12, 13, 14, 15, 16, 17]	"[{'full_name': 'model.layers.11.mlp.experts', 'layer_idx': 11, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	59,454,259,200	47,563,522,048	1.249997	null	null	null	null	null	null	92,196,941,312	80,306,204,160	1.148067	325,369,979,904	313,479,242,752	1.037931	334,636,114,944	322,745,377,792	1.036842	47,563,522,048	322,745,377,792	1.249997	1.036842	null	null	null	null	[]	0.21434	0.146484	"[{'module': 'model.layers.11.mlp.experts', 'fro_rel_error': 0.21556936913131428, 'max_abs_error': 0(...TRUNCATED)	0.005269	0.326961	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	522.425293	259.172821	522.425293	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_11_17', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	59,454,259,200	47,563,522,048	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.11.mlp.experts', 'layer_idx': 11, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.11.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.9885494709014893, 'median': 0.9889452457427979, 'std':(...TRUNCATED)	"{11: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	0.988549	0.988945	0.009027	716,800	7.626105	8.527329	4.154167	716,800	2,245.275635	1,880.783081	1,582.497925	716,800	2,146.422607	1,788.724854	1,561.331665	716,800	0.956048	0.960173	0.037673	716,800	354.265839	316.763611	267.931763	716,800	0.13944	0.156079	0.074526	716,800	0.866599	0.909801	0.153512	716,800	25.307266	24.5222	17.000677	716,800	522.276978	457.172882	252.000961	716,800	581.588379	477.247467	321.568054	716,800	1.093838	1.047102	0.214311	716,800	283.423157	257.494659	233.162094	716,800	0.495203	0.474001	0.343676	716,800	0.989543	0.989912	0.008594	716,800	7.262069	8.145229	4.015381	716,800	2,239.647217	1,890.061279	1,613.1875	716,800	2,307.458252	1,906.134766	1,672.304199	716,800	1.020994	1.011414	0.035246	716,800	347.800232	274.319824	284.920349	716,800	0.132869	0.147127	0.07491	716,800	0.914786	0.931052	0.064666	716,800	22.494347	21.400614	8.172878	716,800	937.599487	832.087158	671.159058	716,800	770.786865	697.302734	593.582153	716,800	0.818485	0.82827	0.114407	716,800	378.429138	335.047913	216.720779	716,800	0.407275	0.394441	0.128629	716,800	0.986036	0.987014	0.008742	716,800	9.12565	9.243566	2.949987	716,800	2,832.042236	2,443.388184	1,765.748901	716,800	2,719.310547	2,327.588379	1,760.718384	716,800	0.953412	0.958512	0.036819	716,800	491.676422	415.337952	304.774536	716,800	0.165733	0.167487	0.051493	716,800	0.931503	0.942996	0.046935	716,800	20.450552	19.439087	6.277029	716,800	1,017.244446	895.421631	752.767822	716,800	1,011.848755	904.627319	731.408813	716,800	0.990756	0.989295	0.12272	716,800	352.765076	316.711151	199.367676	716,800	0.367989	0.346866	0.121522	716,800	0.866561	0.909773	0.153515	716,800	25.313391	24.526047	16.999481	716,800	522.289124	457.180359	252.005875	716,800	581.596375	477.252655	321.568329	716,800	1.09383	1.047061	0.214303	716,800	283.472229	257.549866	233.171997	716,800	0.495303	0.474161	0.343645	716,800	0.914752	0.930993	0.064673	716,800	22.499754	21.409914	8.172344	716,800	937.603333	832.053955	671.162231	716,800	770.792114	697.276978	593.582458	716,800	0.81849	0.828252	0.114426	716,800	378.49173	335.107666	216.747025	716,800	0.407347	0.394554	0.128632	716,800
gpt_oss_20b_mobe_middle_out_layers_11_18	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	"['model.layers.11.mlp.experts', 'model.layers.12.mlp.experts', 'model.layers.13.mlp.experts', 'mode(...TRUNCATED)	[11, 12, 13, 14, 15, 16, 17, 18]	"[{'full_name': 'model.layers.11.mlp.experts', 'layer_idx': 11, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	67,947,724,800	54,358,310,912	1.249997	null	null	null	null	null	null	105,367,932,928	91,778,519,040	1.148067	325,369,979,904	311,780,566,016	1.043586	334,636,114,944	321,046,701,056	1.042328	54,358,310,912	321,046,701,056	1.249997	1.042328	null	null	null	null	[]	0.21355	0.146484	"[{'module': 'model.layers.11.mlp.experts', 'fro_rel_error': 0.21556936913131428, 'max_abs_error': 0(...TRUNCATED)	0.004611	0.329585	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	543.600403	269.283356	543.600403	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_11_18', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	67,947,724,800	54,358,310,912	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.11.mlp.experts', 'layer_idx': 11, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.11.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.987781822681427, 'median': 0.988194465637207, 'std': 0(...TRUNCATED)	"{11: {'cos_before_layer_og_vs_method': {'mean': 1.0000005960464478, 'median': 1.0000005960464478, '(...TRUNCATED)	0.987782	0.988194	0.009391	819,200	7.984946	8.812694	4.089386	819,200	2,624.638184	2,083.836914	1,796.166016	819,200	2,525.997559	1,986.478516	1,780.776123	819,200	0.959235	0.963074	0.037731	819,200	430.216888	362.193024	328.203491	819,200	0.145016	0.160044	0.072968	819,200	0.876023	0.918894	0.14657	819,200	24.466719	23.235096	16.218618	819,200	604.737915	499.798492	328.085236	819,200	677.384216	542.993652	402.555756	819,200	1.099769	1.067698	0.204471	819,200	303.127533	288.348511	229.084854	819,200	0.481182	0.452451	0.326978	819,200	0.989138	0.989584	0.008774	819,200	7.505677	8.276923	3.898544	819,200	2,682.379395	2,070.504395	1,920.338623	819,200	2,756.969971	2,086.791504	1,976.471436	819,200	1.020972	1.013386	0.034224	819,200	422.125977	349.346924	338.086639	819,200	0.136914	0.149554	0.072483	819,200	0.916413	0.931618	0.062497	819,200	22.318047	21.311523	7.959263	819,200	1,135.106445	919.617676	832.123657	819,200	946.566956	777.434326	736.099548	819,200	0.824778	0.834298	0.111415	819,200	447.4646	379.254669	284.021912	819,200	0.403309	0.390326	0.125042	819,200	0.985731	0.986846	0.009045	819,200	9.241669	9.303582	2.930962	819,200	3,427.245361	2,687.933105	2,301.181641	819,200	3,295.846191	2,563.55957	2,262.036865	819,200	0.954996	0.959819	0.035635	819,200	596.020569	490.988708	408.701263	819,200	0.167158	0.167923	0.050838	819,200	0.935067	0.946291	0.045855	819,200	19.849197	18.863598	6.290082	819,200	1,278.024902	992.542603	995.133179	819,200	1,257.455322	1,001.23938	951.520874	819,200	0.987073	0.983817	0.116698	819,200	413.821045	357.953125	258.468445	819,200	0.356322	0.33548	0.120932	819,200	0.875981	0.918866	0.146572	819,200	24.473545	23.239155	16.217197	819,200	604.751709	499.840302	328.091187	819,200	677.391968	542.983154	402.554382	819,200	1.09976	1.067703	0.204463	819,200	303.198547	288.450714	229.104156	819,200	0.481293	0.452543	0.326945	819,200	0.916379	0.931582	0.062504	819,200	22.323511	21.31715	7.958695	819,200	1,135.11145	919.682007	832.128662	819,200	946.57312	777.380493	736.101624	819,200	0.824782	0.834312	0.111432	819,200	447.545624	379.319641	284.068726	819,200	0.403382	0.390381	0.125045	819,200
gpt_oss_20b_mobe_middle_out_layers_10_18	gpt_oss_20b	gpt_oss_moe	openai/gpt-oss-20b	mobe_gate_up_b16_t1536	0	"['model.layers.10.mlp.experts', 'model.layers.11.mlp.experts', 'model.layers.12.mlp.experts', 'mode(...TRUNCATED)	[10, 11, 12, 13, 14, 15, 16, 17, 18]	"[{'full_name': 'model.layers.10.mlp.experts', 'layer_idx': 10, 'module_name': 'mlp.experts', 'modul(...TRUNCATED)	76,441,190,400	61,153,099,776	1.249997	null	null	null	null	null	null	118,538,924,544	103,250,833,920	1.148067	325,369,979,904	310,081,889,280	1.049303	334,636,114,944	319,348,024,320	1.047873	61,153,099,776	319,348,024,320	1.249997	1.047873	null	null	null	null	[]	0.214128	0.146484	"[{'module': 'model.layers.10.mlp.experts', 'fro_rel_error': 0.21883503684464894, 'max_abs_error': 0(...TRUNCATED)	0.004135	0.300398	null	null	null	true	false	false	true	false	false	null	null	null	null	null	{'param_devices': ['cuda:0'], 'buffer_devices': ['cuda:0'], 'hf_device_map': None}	544.328735	266.715271	544.328735	false	['arc_challenge', 'winogrande', 'piqa', 'hellaswag', 'openbookqa']	cfg.run_lm_eval is False	"{'The theory of tensor train decomposition for neural networks suggests that': 'The theory of tenso(...TRUNCATED)	The theory of tensor train decomposition for neural networks suggests that	"The theory of tensor train decomposition for neural networks suggests that the weight matrices of a(...TRUNCATED)	"['openai/gpt-oss-20b', 'gpt_oss_20b_mobe_middle_out_layers_10_18', 'mobe_gate_up_b16_t1536', 0, '[{(...TRUNCATED)	76,441,190,400	61,153,099,776	1.249997	"{'label': 'mobe_gate_up_b16_t1536', 'plans': [{'target': 'moe_experts', 'decomposition_method': 'mo(...TRUNCATED)	"[{'decomposition_per_module': [{'full_name': 'model.layers.10.mlp.experts', 'layer_idx': 10, 'modul(...TRUNCATED)	true	"[{'module': 'model.layers.10.mlp.experts', 'old_type': 'GptOssExperts', 'new_type': 'GptOssExperts'(...TRUNCATED)	"{'cos_before_layer_og_vs_method': {'mean': 0.9853881597518921, 'median': 0.9862595796585083, 'std':(...TRUNCATED)	"{10: {'cos_before_layer_og_vs_method': {'mean': 1.0000004768371582, 'median': 1.0000004768371582, '(...TRUNCATED)	0.985388	0.98626	0.011085	921,600	8.853731	9.50904	4.233065	921,600	2,440.499268	1,881.059326	1,820.73291	921,600	2,342.453369	1,784.694214	1,805.135742	921,600	0.955153	0.959037	0.039924	921,600	437.614685	359.397766	337.987854	921,600	0.160389	0.171798	0.074788	921,600	0.855044	0.899242	0.153228	921,600	27.233717	25.941444	16.349686	921,600	564.419434	457.183411	330.079926	921,600	627.999939	477.708618	404.766571	921,600	1.088823	1.059671	0.218131	921,600	302.964355	272.311951	224.542221	921,600	0.527174	0.49653	0.326211	921,600	0.986756	0.987744	0.010691	921,600	8.391698	8.979601	4.10522	921,600	2,488.251221	1,890.385742	1,937.97937	921,600	2,557.369629	1,906.252441	1,991.869507	921,600	1.021026	1.014539	0.034896	921,600	430.032959	326.248962	350.408722	921,600	0.152283	0.161985	0.075383	921,600	0.906343	0.924367	0.069664	921,600	23.691793	22.426956	8.338663	921,600	1,052.969482	832.854858	819.196655	921,600	870.75415	695.063721	729.327393	921,600	0.81035	0.816633	0.116787	921,600	432.375214	355.721039	285.096344	921,600	0.427499	0.411108	0.126044	921,600	0.98336	0.984934	0.010724	921,600	10.027436	9.958396	3.022018	921,600	3,176.752197	2,443.791016	2,320.367432	921,600	3,049.435303	2,327.976563	2,281.758545	921,600	0.95136	0.956188	0.037869	921,600	591.799988	461.568207	423.178864	921,600	0.181024	0.178856	0.051565	921,600	0.921068	0.934953	0.054498	921,600	21.906784	20.779461	6.99464	921,600	1,177.22583	896.15686	980.989502	921,600	1,161.323975	911.486694	939.759583	921,600	0.995292	0.994687	0.125359	921,600	408.15036	340.796204	262.901642	921,600	0.393499	0.372602	0.129467	921,600	0.855003	0.899212	0.153229	921,600	27.239645	25.94532	16.348452	921,600	564.43219	457.186523	330.086151	921,600	628.009277	477.696136	404.77066	921,600	1.088816	1.059689	0.218125	921,600	303.02832	272.402405	224.568527	921,600	0.52727	0.496663	0.326185	921,600	0.906307	0.924339	0.069675	921,600	23.696934	22.431122	8.338612	921,600	1,052.973999	832.922546	819.201477	921,600	870.760742	695.053406	729.32843	921,600	0.810356	0.816622	0.116801	921,600	432.447418	355.728455	285.146851	921,600	0.427565	0.411176	0.126052	921,600

End of preview.

MoBE factor cache — GPT-OSS-20B (num_B=16)

Trained MoBE (Mixture-of-Basis-Experts) factors for the MoE gate_proj / up_proj matrices of GPT-OSS-20B, plus the perplexity / activation-geometry results of the compression sweep that produced them.

mobe_factor_cache/<model>/layer_<i>/<projection>__nB<num_B>_t<T>_<activation>_E<E>_R<R>_C<C>.pt Each file is a torch.save payload with factors (A, B, W), target_std, meta (hyperparameters, final loss, best epoch, training time) and the loss history. Reconstruct with reconstruct_from_factors(A, B, W, activation, target_std).
*_results.csv / *_results.json — perplexity (WikiText-2, C4), compression ratios and residual-stream activation geometry for each cumulative set of compressed MoE layers.
*__partial_results* — per-target-set checkpoints written during the sweep.

Configuration: num_B=16, activation tanh, 10000 epochs per projection.

These are derived caches, not a standalone model.

Downloads last month: 65

Total file size:

20.4 GB

MoBE factor cache — GPT-OSS-20B (num_B=16)

Contents