You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 
zhanghangit bf716fb7c0 add nvidia megatron code 3 years ago
..
.ipynb_checkpoints add nvidia megatron code 3 years ago
data add nvidia megatron code 3 years ago
deprecated_data_utils add nvidia megatron code 3 years ago
fp16 add nvidia megatron code 3 years ago
fp16_cpm add nvidia megatron code 3 years ago
fused_kernels add nvidia megatron code 3 years ago
model add nvidia megatron code 3 years ago
mpu add nvidia megatron code 3 years ago
tokenizer add nvidia megatron code 3 years ago
__init__.py add nvidia megatron code 3 years ago
arguments.py add nvidia megatron code 3 years ago
checkpointing.py add nvidia megatron code 3 years ago
eval_ppl.py add nvidia megatron code 3 years ago
global_vars.py add nvidia megatron code 3 years ago
indexer.py add nvidia megatron code 3 years ago
initialize.py add nvidia megatron code 3 years ago
learning_rates.py add nvidia megatron code 3 years ago
memory.py add nvidia megatron code 3 years ago
module.py add nvidia megatron code 3 years ago
package_info.py add nvidia megatron code 3 years ago
text_generation_utils.py add nvidia megatron code 3 years ago
training.py add nvidia megatron code 3 years ago
utils.py add nvidia megatron code 3 years ago

使用100G中文高质量语料,128张V100,训练的中文Megatron模型,参数量2.6B

Python C++ Shell Cuda TeX other

Contributors (2)