|
{ |
|
"auto_loss_scaling": false, |
|
"device_iterations": 32, |
|
"embedding_serialization_factor": 1, |
|
"enable_half_partials": true, |
|
"executable_cache_dir": "/tmp/exe_cache/packed_bert_squad/", |
|
"execute_encoder_on_cpu_for_generation": false, |
|
"gradient_accumulation_steps": 32, |
|
"inference_device_iterations": 64, |
|
"inference_replication_factor": 1, |
|
"ipus_per_replica": 4, |
|
"layers_per_ipu": [ |
|
0, |
|
4, |
|
4, |
|
4 |
|
], |
|
"matmul_proportion": 0.22, |
|
"optimizer_state_offchip": false, |
|
"optimum_version": "1.6.1", |
|
"output_mode": "final", |
|
"recompute_checkpoint_every_layer": true, |
|
"replicated_tensor_sharding": false, |
|
"replication_factor": 1, |
|
"seed": 42, |
|
"transformers_version": "4.25.1" |
|
} |
|
|