关于GCU、沐曦GPGPU、MLU、0卡V100资源4月7日恢复上架的公告>>> 关于共建具身智能开源数据集的倡议>>> 关于云脑任务中统一路径访问方式的公告>>> 关于将启智集群GPU资源迁移至智算集群的公告>>>

imyzx 10f59c5675 'modify for 200B-serving'		4 months ago
scripts	'update for backend independent with inference_engine'	4 months ago

search	raw_ma_910B_pcmind200b	5 months ago

src	raw_ma_910B_pcmind200b	5 months ago

strategy	raw_ma_910B_pcmind200b	5 months ago

tokenizer	'update'	4 months ago

whls	'add frontend'	4 months ago

README.md	raw_ma_910B_pcmind200b	5 months ago

client_api.py	'modify 1227'	4 months ago

config.py	'add config.py'	4 months ago

dump.json	raw_ma_910B_pcmind200b	5 months ago

export_to_mindir_200B.py	raw_ma_910B_pcmind200b	5 months ago

hccn_net_health.sh	raw_ma_910B_pcmind200b	5 months ago

hccn_tool_config.sh	raw_ma_910B_pcmind200b	5 months ago

hidden_export.py	raw_ma_910B_pcmind200b	5 months ago

ma-pre-start.sh	raw_ma_910B_pcmind200b	5 months ago

npu_ip_set.sh	raw_ma_910B_pcmind200b	5 months ago

parallel_speed_up.json	raw_ma_910B_pcmind200b	5 months ago

paramter_convert.py	raw_ma_910B_pcmind200b	5 months ago

paramter_divide.py	raw_ma_910B_pcmind200b	5 months ago

paramter_merge.py	raw_ma_910B_pcmind200b	5 months ago

predict_200B.py	raw_ma_910B_pcmind200b	5 months ago

predict_200B_v1.py	raw_ma_910B_pcmind200b	5 months ago

predict_200B_v2.py	raw_ma_910B_pcmind200b	5 months ago

predict_200B_v2_localNPU.py	'modify 1227'	4 months ago

predict_200B_v2_localNPU_exportMindIR.py	'update for pcmind200B_v2 mindIR generation'	4 months ago

rank_table_8pcs.json	'update npu-mem'	4 months ago

rank_table_16pcs.json	raw_ma_910B_pcmind200b	5 months ago

run_pridect.sh	raw_ma_910B_pcmind200b	5 months ago

serving_200B.py	'modify 1227 - fix03'	4 months ago

serving_200B_mindir_910B_8cards.py	'update for pcmind200B_v2 mindIR loading'	4 months ago

serving_local_npu.py	'modify for 200B-serving'	4 months ago

start.sh	raw_ma_910B_pcmind200b	5 months ago

strategy_export.py	raw_ma_910B_pcmind200b	5 months ago

train_taoht_200b-exp00.py	raw_ma_910B_pcmind200b	5 months ago

PengCheng.Mind-200B

generate pipeline configuration

Set the dp(optimizer shard), tp, pp, layer_num, etc. in search/searcher.py.
Generate the configuration file under BaSys_FineGrain.

python3 searcher.py

run the training with pipeline_config_filename={WORK_DIR}/search/BaSys_FineGrain/{strategy_name}.json

No Description

Python Shell

How to access data resources in code

README.md

PengCheng.Mind-200B

generate pipeline configuration

Contributors (1) All

Contributors (1)
All