Skip to content

[3/N] add prefetch support for CUDA backend : running ds4 for any GPU with cache (2.75 x faster!)#402

Open
yiakwy-xpu-ml-framework-team wants to merge 1 commit into
antirez:mainfrom
yiakwy-xpu-ml-framework-team:add_prefetch_cache_support_for_cuda
Open

[3/N] add prefetch support for CUDA backend : running ds4 for any GPU with cache (2.75 x faster!)#402
yiakwy-xpu-ml-framework-team wants to merge 1 commit into
antirez:mainfrom
yiakwy-xpu-ml-framework-team:add_prefetch_cache_support_for_cuda

Conversation

@yiakwy-xpu-ml-framework-team

@yiakwy-xpu-ml-framework-team yiakwy-xpu-ml-framework-team commented Jun 12, 2026 •
edited
Loading

Copy link
Copy Markdown

Introduction

We have verified our sft/rl (much stronger dsv4) with 2 bits can run in 15 tokens/sec.

Then it came into my mind, if I run it in other GPU with UVM technologies (Mapping CPU memory to GPU memory) with prefetch cache ?

For example , we have 80 GB GPU, and we preload tensors with 64 GB from 154 GB model.

Now we did it.

Server side snapshot
截屏2026-06-13 02 58 25
截屏2026-06-13 02 58 15

** client side snapshot **
087e126d257ee6806d4d640230a6e6c6

Acceleration

Config Speed (tokens/sec) Model (GB)
dsv4 iq2_xxs 15 81
q4 2 154
q4 + 64GB cache 5.5 154

Discussion

This is follow up of #368 and #377, but we can merge it independently since it works on runtime engine not quantization toolkits.

@yiakwy-xpu-ml-framework-team

yiakwy-xpu-ml-framework-team commented Jun 12, 2026

Copy link
Copy Markdown
Author

@antirez Sorry for disturbing you again! But this is a real important feature with prefetch cache support in CUDA backend (any cuda) !

@yiakwy-xpu-ml-framework-team

yiakwy-xpu-ml-framework-team commented Jun 12, 2026

Copy link
Copy Markdown
Author

How to use it:

DS4_SFT_E2_FP4_MODEL=./gguf/DeepSeek-V4-Flash_e2_v1_Q4KExperts-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix.gguf

# ds4: --ssd-streaming is currently supported only with --metal
#   --ssd-streaming \

# export DS4_CUDA_MODEL_PRELOAD_SIZE_GB=64
export DS4_CUDA_WEIGHT_CACHE_LIMIT_GB=64

# export DS4_CUDA_WEIGHT_PRELOAD=1
export DS4_CUDA_WEIGHT_CACHE=1

# for debugging
export DS4_CUDA_MODEL_COPY_VERBOSE=1

export DS4_CUDA_WEIGHT_CACHE_VERBOSE=1

# important !
export DS4_CUDA_COPY_MODEL_CHUNKED=1

CUDA_VISIBLE_DEVICES=2 \
DS4_MODEL_NAME="deepseek-v4-flash-rl-e5" \
DS4_LOCK_FILE=/tmp/ds4-server-2.lock ./ds4-server \
  --cuda \
  -m $DS4_SFT_E2_FP4_MODEL \
  --ssd-streaming-cache-experts 64GB \
  --ctx 256000 \
  --kv-disk-dir /raid/yiakwy/tmp/ds4-kv-gpu4 \
  --kv-disk-space-mb 102400 \
  --host 127.0.0.1 \
  --port 8001

@yiakwy-xpu-ml-framework-team yiakwy-xpu-ml-framework-team changed the title [3/N] add prefetch support for CUDA backend : running ds4 for any GPU with cache (3x faster!) [3/N] add prefetch support for CUDA backend : running ds4 for any GPU with cache (2.75 x faster!) Jun 12, 2026
This was referenced Jun 12, 2026
Open
Open
Open
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@yiakwy-xpu-ml-framework-team