Skip to content
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

Jump to bottom

estimate_zero2_model_states_mem_needs: fixing memory estiamtion #5099

Merged
merged 6 commits into from
Jun 4, 2024
Merged

estimate_zero2_model_states_mem_needs: fixing memory estiamtion #5099

Changes from 2 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
13fbabb
estimate_zero2_model_states_mem_needs: fixing memory estiamtion
nelyahu Feb 8, 2024
7a1e63d
Merge branch 'master' into fix_memory_estimation
tjruwase Feb 9, 2024
3b7a053
Update deepspeed/runtime/zero/stage_1_and_2.py
nelyahu Apr 22, 2024
9b1b7e0
Update deepspeed/runtime/zero/stage_1_and_2.py
nelyahu Apr 22, 2024
cd57043
Merge branch 'master' into fix_memory_estimation
tjruwase May 13, 2024
52f6c39
Merge branch 'master' into fix_memory_estimation
tjruwase May 28, 2024
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
4 changes: 3 additions & 1 deletion deepspeed/runtime/zero/stage_1_and_2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2427,7 +2427,9 @@ def estimate_zero2_model_states_mem_needs(total_params,
gpu_mem = 2 * total_params
cpu_mem = total_params * max(4 * total_gpus, 16) * additional_buffer_factor
else:
gpu_mem = 4 * total_params + int(16 * total_params / total_gpus)
# GPU's total_params multipliers: 2 = params_16bit,
# 14 = 2_grads_16bit + 4_params_32bit + 8_optimizer_states_32bit(momentum and variance)
tjruwase marked this conversation as resolved.
Show resolved Hide resolved
nelyahu marked this conversation as resolved.
Show resolved Hide resolved
gpu_mem = 2 * total_params + int(14 * total_params / total_gpus)
nelyahu marked this conversation as resolved.
Show resolved Hide resolved
cpu_mem = total_params * 4 * num_gpus_per_node * additional_buffer_factor

return int(cpu_mem), int(gpu_mem)
Expand Down
Loading