Skip to content

[https://nvbugs/6826000][fix] Size the DeepSeek-V4-Pro AgentX ctx KV pools for the compressed history #20970

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

Closed
chenfeiz0326 wants to merge 1 commit into from
Closed

[https://nvbugs/6826000][fix] Size the DeepSeek-V4-Pro AgentX ctx KV pools for the compressed history #20970

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 ...300_deepseek-v4-pro-dspark_agentx_con1156_ctx2_dep8_gen1_dep8_eplb0_dspark3_ccb-NIXL.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,11 @@ worker_config:
block_reuse_config:
policy: per_conversation
max_num_turns: 5
pool_ratio: [0.55, 0.22, 0.23]
# DeepSeek-V4 KV cache manager v2 life cycles, in order: SWA windows, the compressed
# history with the indexer keys, compressor state windows. With per_conversation
# reuse only the history grows with the retained conversations, so it takes most of
# the GPU and host pools.
pool_ratio: [0.18, 0.64, 0.18]
host_cache_size: 193273528320
moe_config:
backend: CUTEDSL
Expand Down
6 changes: 5 additions & 1 deletion ...00_deepseek-v4-pro-dspark_agentx_con1456_ctx3_dep8_gen1_dep16_eplb0_dspark5_ccb-NIXL.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,11 @@ worker_config:
block_reuse_config:
policy: per_conversation
max_num_turns: 5
pool_ratio: [0.55, 0.22, 0.23]
# DeepSeek-V4 KV cache manager v2 life cycles, in order: SWA windows, the compressed
# history with the indexer keys, compressor state windows. With per_conversation
# reuse only the history grows with the retained conversations, so it takes most of
# the GPU and host pools.
pool_ratio: [0.18, 0.64, 0.18]
host_cache_size: 193273528320
moe_config:
backend: CUTEDSL
Expand Down
Loading