{ | |
"optimizer_state_offchip": true, | |
"replicated_tensor_sharding": true, | |
"enable_half_partials": true, | |
"gradient_accumulation_steps": 16, | |
"executable_cache_dir": "./exe_cache", | |
"layers_per_ipu": [4, 4], | |
"matmul_proportion": 0.6, | |
"projection_serialization_factor": 4 | |
} |