..
|
data_efficiency
|
mii
|
zeropp
|
1.3B-MoE-128.png
|
175b-trend.png
|
1cycle_lr.png
|
1t-trend.png
|
3d-parallelism.png
|
530b-trend.png
|
DeepSpeed-pillars.png
|
DeepSpeed-vs-Megatron.png
|
DeepSpeed_dark.svg
|
DeepSpeed_dark_transparent.svg
|
DeepSpeed_light.svg
|
DeepSpeed_light_transparent.svg
|
accelerate-dark.png
|
accelerate-light.png
|
accelerate.png
|
adam-convergence.png
|
auto-tp-chart-latency.png
|
auto-tp-chart-opt-throughput.png
|
auto-tp-chart-throughput.png
|
bert-ib.png
|
bert-large-training-time.png
|
bert-scaling.png
|
bert-tcp.png
|
bingbert-mixedbit.png
|
convergence-table.png
|
deepspeed-logo-uppercase-bold-white-1.15.svg
|
deepspeed-logo-uppercase-bold-white.svg
|
deepspeed-logo-uppercase-white.svg
|
deepspeed-speedup.png
|
determined.svg
|
end-to-end-bert-training.PNG
|
evoformer.png
|
gpu-numbers.png
|
hf-logo.png
|
hf-transformers.png
|
inference-gemm-scheduling.png
|
inference-kernel-fusion.png
|
inference-latency.png
|
inference-throughput.png
|
large-model-graph.png
|
layernorm_animation.gif
|
layernorm_deepspeed.gif
|
layernorm_ds.png
|
layernorm_pytorch.gif
|
layernorm_torch.png
|
lightning-dark.png
|
lightning-dark.svg
|
lightning-light.svg
|
lightning.png
|
loss_and_lr.png
|
lr_schedule.png
|
megatron-gpt2-perf-test.png
|
model_convergence.png
|
moe-lat-tput.png
|
moe-nlg.png
|
mosaicml.svg
|
new-megatron-ds.png
|
old-vs-new-azure.png
|
onebit-adam-overview.png
|
onebit-convergence.png
|
oom_dp8_1.5B_log.png
|
perf-overview.png
|
pipe-schedule.png
|
pp-lowbw-gpt2.png
|
prmoe.png
|
qkv_fusion.png
|
quantization-8bit.png
|
quantization-mixedbit.png
|
sa_backward_pass.png
|
sa_bert_base_time_result.png
|
sa_bert_large_time_result.png
|
sa_fixed_sparsity_structure.png
|
sa_forward_pass.png
|
sa_gpt2_time_result.png
|
sa_long_document_comprehension_result.png
|
sa_maximum_sequence_runnable_on_bert.png
|
sa_variable_sparsity_structure.png
|
softmax_animation.gif
|
softmax_deepspeed.gif
|
softmax_ds.png
|
softmax_pytorch.gif
|
softmax_torch.png
|
squad-ib.png
|
squad-scaling.png
|
squad-tcp.png
|
tensorboard_monitor.PNG
|
transformer_kernel_perf.png
|
transformer_kernel_perf_seq128.PNG
|
transformer_kernel_perf_seq512.PNG
|
transformer_preln_arch.png
|
transformers-dark.png
|
transformers-light.png
|
variable_sparsity_pattern.png
|
vl_moe.png
|
vmss-setup.png
|
wandb_monitor.PNG
|
webinar-aug2020.png
|
xtc-1.png
|
xtc-2.png
|
xtc-3.png
|
xtc-4.png
|
zero-full.png
|
zero1_dp8_1.5B_log.png
|
zero1_dp8_1.5B_smi.png
|
zero2_dp32_10B_log.png
|
zero2_dp32_10B_smi.png
|
zero3-offload-1-v100.png
|
zero3-offload-16-v100.png
|
zero3-offload-200B-scalability.png
|
zero3-offload-512-v100.png
|
zero3-offload-memory-overview.png
|
zero_inference_full_offload.png
|
zero_inference_model_scale.png
|
zero_inference_models.png
|
zero_inference_multi_gpu.png
|
zero_inference_prefetch.png
|
zero_inference_token_count_batch_size.png
|
zero_inference_token_count_cpu_throughput.png
|
zero_inference_token_count_nvme_throughput.png
|
zero_offload_dp1_10B_cpu.png
|
zero_offload_dp1_10B_log.png
|
zero_offload_dp1_10B_smi.png
|