mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2026-05-02 21:51:30 +00:00
57d14d22bc
* refactor: move legacy code to archive/ directory - Moved ktransformers, csrc, third_party, merge_tensors to archive/ - Moved build scripts and configurations to archive/ - Kept kt-kernel, KT-SFT, doc, and README files in root - Preserved complete git history for all moved files * refactor: restructure repository to focus on kt-kernel and KT-SFT modules * fix README * fix README * fix README * fix README * docs: add performance benchmarks to kt-kernel section Add comprehensive performance data for kt-kernel to match KT-SFT's presentation: - AMX kernel optimization: 21.3 TFLOPS (3.9× faster than PyTorch) - Prefill phase: up to 20× speedup vs baseline - Decode phase: up to 4× speedup - NUMA optimization: up to 63% throughput improvement - Multi-GPU (8×L20): 227.85 tokens/s total throughput with DeepSeek-R1 FP8 Source: https://lmsys.org/blog/2025-10-22-KTransformers/ This provides users with concrete performance metrics for both core modules, making it easier to understand the capabilities of each component. * refactor: improve kt-kernel performance data with specific hardware and models Replace generic performance descriptions with concrete benchmarks: - Specify exact hardware: 8×L20 GPU + Xeon Gold 6454S, Single/Dual-socket Xeon + AMX - Include specific models: DeepSeek-R1-0528 (FP8), DeepSeek-V3 (671B) - Show detailed metrics: total throughput, output throughput, concurrency details - Match KT-SFT presentation style for consistency This provides users with actionable performance data they can use to evaluate hardware requirements and expected performance for their use cases. * fix README * docs: clean up performance table and improve formatting * add pic for README * refactor: simplify .gitmodules and backup legacy submodules - Remove 7 legacy submodules from root .gitmodules (archive/third_party/*) - Keep only 2 active submodules for kt-kernel (llama.cpp, pybind11) - Backup complete .gitmodules to archive/.gitmodules - Add documentation in archive/README.md for researchers who need legacy submodules This reduces initial clone size by ~500MB and avoids downloading unused dependencies. * refactor: move doc/ back to root directory Keep documentation in root for easier access and maintenance. * refactor: consolidate all images to doc/assets/ - Move kt-kernel/assets/heterogeneous_computing.png to doc/assets/ - Remove KT-SFT/assets/ (images already in doc/assets/) - Update KT-SFT/README.md image references to ../doc/assets/ - Eliminates ~7.9MB image duplication - Centralizes all documentation assets in one location * fix pic path for README
62 lines
1.8 KiB
Python
62 lines
1.8 KiB
Python
#!/usr/bin/env python
|
|
# coding=utf-8
|
|
'''
|
|
Description :
|
|
Author : chenht2022
|
|
Date : 2024-07-25 10:32:05
|
|
Version : 1.0.0
|
|
LastEditors : chenht2022
|
|
LastEditTime : 2024-08-06 10:36:59
|
|
Copyright (c) 2024 by KVCache.AI, All Rights Reserved.
|
|
'''
|
|
import os, sys
|
|
import time
|
|
sys.path.append(os.path.dirname(__file__) + '/../build')
|
|
import cpuinfer_ext
|
|
import torch
|
|
|
|
input_size = 16384
|
|
output_size = 5120
|
|
stride = 32
|
|
group_max_len = 1024
|
|
proj_type = 1 # ggml_type::GGML_TYPE_F16
|
|
hidden_type = 1 # ggml_type::GGML_TYPE_F16
|
|
qlen = 30
|
|
layer_num = 10
|
|
CPUInfer = cpuinfer_ext.CPUInfer(48)
|
|
validation_iter = 100
|
|
|
|
with torch.inference_mode(mode=True):
|
|
linears = []
|
|
projs = []
|
|
for _ in range(layer_num):
|
|
proj = torch.randn((output_size, input_size), dtype=torch.float16, device = "cuda").to("cpu").contiguous()
|
|
config = cpuinfer_ext.linear.LinearConfig(input_size, output_size, stride, group_max_len, proj.data_ptr(), proj_type, hidden_type)
|
|
linear = cpuinfer_ext.linear.Linear(config)
|
|
projs.append(proj)
|
|
linears.append(linear)
|
|
|
|
# validation
|
|
for i in range(validation_iter):
|
|
linear = linears[i % layer_num]
|
|
input = torch.randn((qlen, input_size), dtype=torch.float16).contiguous()
|
|
output = torch.empty((qlen, output_size), dtype=torch.float16).contiguous()
|
|
input = input / 100
|
|
|
|
CPUInfer.submit(
|
|
linear.forward(
|
|
qlen,
|
|
input.data_ptr(),
|
|
output.data_ptr()
|
|
)
|
|
)
|
|
CPUInfer.sync()
|
|
# print('cpuinfer output', output)
|
|
|
|
proj = projs[i%layer_num]
|
|
t_output = torch.mm(input, proj.t())
|
|
# print('torch output', t_output)
|
|
|
|
diff = torch.mean(torch.abs(output - t_output)) / torch.mean(torch.abs(t_output))
|
|
print('diff = ', diff)
|
|
assert(diff < 0.001)
|