Name
..
1f1b.gif
1f1b.png
GPT-X.png
adam.png
allgather.png
allreduce.png
allreduce_1.png
allreduce_2.png
apex.png
arch_is_not_important.png
backends.png
backward_step.png
big_science.png
broadcast.png
checkpoint_1.png
checkpoint_2.png
checkpoint_3.png
checkpoint_act.gif
checkpoint_full_act.gif
checkpoint_no_act.gif
cl_1.png
collective.png
column_parallel.png
cpu_adam.png
data_parallelism.png
ddp.png
ddp_analysis_1.png
ddp_analysis_2.png
ddp_analysis_3.png
ddp_analysis_4.png
ddp_analysis_5.png
dp_backward.png
dp_forward.png
dp_forward_2.png
dp_training.png
eleuther_ai.png
gather.png
gpipe_1.png
gpipe_2.png
gpipe_3.gif
hard_core_engineering.png
inter_layer.png
inter_layer_2.png
inter_layer_3.gif
interleaved_1.png
intra_layer.png
intra_layer_2.png
is_large_scale_easy.png
kernel_fusion.png
lm_head.png
max_pooling.png
megatron_3d.png
megatron_attention.jpeg
megatron_block.png
megatron_lm.jpeg
megatron_mlp.png
megatron_mlp_2.png
megatron_mlp_3.png
megatron_mlp_4.png
megatron_mlp_5.png
megatron_mlp_6.png
memory.png
message_passing.png
mixed_precision_1.png
mixed_precision_2.png
mixed_precision_33.png
mixed_precision_4.png
mixed_precision_5.png
model_parallelism.png
multi_head_attention.png
multi_process_1.png
multi_process_2.png
one_bit_adam.png
open_mpi.png
p2p.png
parallelformers.png
parallelism.png
pipe_dream.png
pipe_dream_2bw.png
pipe_dream_flush.png
pipe_dream_flush_2.png
pipeline_flush.png
pipeline_parallelism.png
pipeline_parallelism2.png
pld_1.png
pld_2.png
pld_3.png
process_terms.png
rank_error.png
reduce.png
reduce_scatter.png
ring_allreduce.gif
ring_allreduce.png
row_parallelism.png
scale_is_all_you_need.png
scatter.png
tensor_replace.png
vpce_1.png
vpce_2.png
vpce_3.png
vpe_1.png
vpe_2.png
which_backend.png
why_large_scale.png
zero_1.png
zero_2.png
zero_3.png
zero_infinity.mp4
zero_infinity.png
zero_off_1.png
zero_off_2.png
zero_off_3.png
zero_off_4.png
zero_r_1.png
zero_r_2.png
zero_r_3.jpeg
zero_video.mp4