doremi-llama-280m-proxy / 100000 /optimizer /optimizer_config.json
neuralink's picture
neuralink HF staff
upload the 100k checkpoinmt
fc22efd
raw
history blame
125 Bytes
{"type": "OptimizerFromGradientAccumulator", "parallelism": {"tp_size": "2", "dp_size": "16", "pp_size": "1"}, "configs": {}}