kvcache-ai-ktransformers/doc/en/DeepseekR1_V3_tutorial.md

Report

Prerequisites

We run our best performance tests on
cpu: Intel(R) Xeon(R) Gold 6454S 1T DRAM(2 NUMA nodes)
gpu: 4090D 24G VRAM

Bench result

V0.2

settings

• model: DeepseekV3-q4km（int4）
  
• CPU: cpu_model_name：Intel(R) Xeon(R) Gold 6454S, 32 cores per socket, 2 socket, 2numa nodes
• GPU: 4090D 24GVRAM
• we test after enough warm up!

memory consumption:

• single socket: 382G DRAM, 12G VRAM
• dual socket: 1T DRAM, 12G VRAM

Benchmark Results

"6 experts" case is part of v0.3's preview

Prompt (500 tokens)	Dual socket Ktrans (6 experts)	Dual socket Ktrans (8 experts)	Single socket Ktrans (6 experts)	Single socket Ktrans (8 experts)	llama.cpp (8 experts)
Prefill token/s	97.32	82.94	65.14	54.21	10.31
Decode token/s	13.69	12.208	10.303	8.73	4.51

The highest speedup reaches up to 3.03x in decoding and 9.44x in prefill.

V0.3-Preview

settings

• model: DeepseekV3-BF16 (online quant into int8 for CPU and int4 for GPU)
• CPU: cpu_model_name：Intel(R) Xeon(R) Gold 6454S, 32 cores per socket, 2 socket, 2 numa nodes
• GPU: (1~4)x 4090D 24GVRAM (requires more VRAM for longer prompt)

memory consumptions:

• 644GB DRAM, at least 12GB VRAM

Benchmark Results

Prompt length	1K	2K	4K	8K
KTrans (8 experts) Prefill token/s	185.96	255.26	252.58	195.62
KTrans (6 experts) Prefill token/s	203.70	286.55	271.08	207.20

The prefill of KTrans V0.3 is up to 3.45x times faster than KTrans V0.2, and is up to 63.53x times faster than llama.cpp. The decoding speed is the same as KTrans V0.2 (6 experts version) so it is omitted.

The main acceleration comes from

• Intel AMX instruction set and our specially designed cache friendly memory layout
• Expert selection strategy that selects fewer experts based on offline profile results of out of domain data

how to run

v0.2 showcase

single socket version(32 cores)

our local_chat test command is:

git clone https://github.com/kvcache-ai/ktransformers.git
cd ktransformers
numactl -N 1 -m 1 python ./ktransformers/local_chat.py --model_path <your model path> --gguf_path <your gguf path>  --prompt_file <your promt txt file>  --cpu_infer 33  --cache_lens 1536 
<when you see chat, then press enter to load the text prompt_file>

<your model path> can be local or set from onlie hugging face like deepseek-ai/DeepSeek-V3. If onlie encounters connection problem, try use mirror(hf-mirror.com)
<your gguf path> can also be onlie, but as its large we recommend you download it and quantize the model to what you want.
the command numactl -N 1 -m 1 aims to adoid data transfer between numa nodes.

dual socket version(64 cores)

make suer before you install(use install.sh or make dev_install), setting the env var USE_NUMA=1 by export USE_NUMA=1(if already installed, reinstall it with this env var set)
our local_chat test command is:

git clone https://github.com/kvcache-ai/ktransformers.git
cd ktransformers
export USE_NUMA=1
make dev_install # or sh ./install.sh
python ./ktransformers/local_chat.py --model_path <your model path> --gguf_path <your gguf path>  --prompt_file <your promt txt file>  --cpu_infer 65  --cache_lens 1536 
<when you see chat, then press enter to load the text prompt_file>

The parameters meaning is the same. But As we use dual socket, so we set cpu_infer to 65.

some explanations

1. Also we want to make further use of our two NUMA nodes on Xeon Gold cpu. To avoid the cost of data transfer between nodes, we "copy" the critical matrix on both nodes which takes more memory consumption but accelerates the prefill and decoding process. But this method takes huge memory and slow when loading weights, So be patient when loading and monitor the memory usage.(we are considering to make this method as an option). We are going to optimize this huge memory overhead. Stay tuned~
   
2. the command args --cpu_infer 65 specifies how many cores to use(it's ok that it exceeds the physical number, but it's not the more the better. Adjust it slightly lower to your actual number of cores)