Skip to content

DeepSeek-V3 recipes for H100 #2197

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
scsudhakaran merged 1 commit into from
Feb 10, 2026
Merged

DeepSeek-V3 recipes for H100 #2197

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion scripts/performance/configs/deepseek/deepseek_llm_pretrain.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -210,7 +210,11 @@ def deepseek_v3_pretrain_config_h100(
cfg.model.pipeline_model_parallel_size = base_cfg.pipeline_model_parallel_size
cfg.model.virtual_pipeline_model_parallel_size = base_cfg.virtual_pipeline_model_parallel_size
cfg.model.moe_flex_dispatcher_backend = base_cfg.moe_flex_dispatcher_backend
cfg.model.pipeline_model_parallel_layout = "Et|(tt|)*30mL"
if base_cfg.pp_layout:
cfg.model.pipeline_model_parallel_layout = base_cfg.pp_layout
else:
# Recompute layout based on updated PP/VP sizes
set_deepseek_v3_pipeline_model_parallel_layout(cfg.model)

set_deepseek_v3_common_configs(cfg)
set_workload_base_configs(cfg, base_cfg)
Expand Down
9 changes: 7 additions & 2 deletions scripts/performance/configs/deepseek/deepseek_workload_base_configs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -120,14 +120,15 @@
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_V1 = replace(
BASE_DEEPSEEK_V3_CONFIG,
num_gpus=1024,
tensor_model_parallel_size=4, # TODO: TP=2 is OOM. Resolve it and revert it to recover perf
tensor_model_parallel_size=2,
pipeline_model_parallel_size=8,
virtual_pipeline_model_parallel_size=4,
expert_model_parallel_size=64,
global_batch_size=8192,
recompute_modules=["mla_up_proj", "mlp"],
moe_flex_dispatcher_backend="hybridep",
moe_a2a_overlap=False,
pp_layout="Et|(tt|)*30mL",
)
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_BF16_V1 = DEEPSEEK_V3_PRETRAIN_CONFIG_H100_V1
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_FP8_CS_V1 = DEEPSEEK_V3_PRETRAIN_CONFIG_H100_V1
Expand Down Expand Up @@ -189,7 +190,11 @@
)
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_BF16_V2 = DEEPSEEK_V3_PRETRAIN_CONFIG_H100_V2
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_FP8_CS_V2 = DEEPSEEK_V3_PRETRAIN_CONFIG_H100_V2
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_FP8_SC_V2 = DEEPSEEK_V3_PRETRAIN_CONFIG_H100_FP8_CS_V2
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_FP8_SC_V2 = replace(
DEEPSEEK_V3_PRETRAIN_CONFIG_H100_FP8_CS_V2,
virtual_pipeline_model_parallel_size=2,
pp_layout=None,
)


# =============================================================================
Expand Down
8 changes: 8 additions & 0 deletions scripts/performance/perf_plugins.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -267,6 +267,14 @@ def _set_model_specific_environment_variables(
):
if compute_dtype in ["fp8_cs", "fp8_mx"]:
executor.env_vars["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"
elif (
model_family_name in ["deepseek"]
and model_recipe_name in ["deepseek_v3"]
and train_task == "pretrain"
and gpu in ["h100"]
):
executor.env_vars["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True"

del_cudnn_ln = True
if gpu in ["h100"]:
if model_family_name == "llama" and model_recipe_name == "llama3_8b" and train_task == "pretrain":
Expand Down