Method autotp_size

deepspeed/runtime/engine.py:1291–1292 · view source on GitHub ↗

(self)

Source from the content-addressed store, hash-verified

1289	return self._config.tensor_parallel_config
1290
1291	def autotp_size(self):
1292	return self._config.tensor_parallel_config.autotp_size
1293
1294	def graph_harvesting(self):
1295	return self._config.graph_harvesting

__init__Method · 0.95

_configure_expert_parallelMethod · 0.95

_configure_tensor_parallel_statesMethod · 0.95

_apply_autotp_partitioningMethod · 0.95

_validate_zero3_moe_compatibilityMethod · 0.95

_consolidated_16bit_state_dictMethod · 0.95

_zero3_consolidated_16bit_state_dictMethod · 0.95

_cpu_folded_runtime_workerFunction · 0.80

test_h100_folded_tp2_ep2_runtimeMethod · 0.80

test_h100_folded_tp2_ep4_runtimeMethod · 0.80

_compare_tp_gradientsMethod · 0.80

test_tp_plan_basic_trainingMethod · 0.80

no outgoing calls

_cpu_folded_runtime_workerFunction · 0.64

test_h100_folded_tp2_ep2_runtimeMethod · 0.64

test_h100_folded_tp2_ep4_runtimeMethod · 0.64

_compare_tp_gradientsMethod · 0.64

test_tp_plan_basic_trainingMethod · 0.64

test_tp_plan_attaches_universal_checkpoint_infoMethod · 0.64

test_tp_plan_with_zero1Method · 0.64

test_tp_plan_with_zero2Method · 0.64

test_qwen2_tp_plan_with_zero2Method · 0.64

test_qwen2_tied_lm_head_falls_back_to_replicatedMethod · 0.64

test_custom_model_with_custom_tp_planMethod · 0.64

test_tp_model_init_merges_configMethod · 0.64