mirror of
https://github.com/kvcache-ai/ktransformers.git
synced 2026-04-28 20:00:06 +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
40 lines
No EOL
1.2 KiB
Python
40 lines
No EOL
1.2 KiB
Python
import os
|
|
os.environ["CUDA_VISIBLE_DEVICES"]="1"
|
|
# add path
|
|
import sys
|
|
sys.path.append("../..")
|
|
import pycuda.autoinit
|
|
import pycuda.driver as cuda
|
|
from pycuda.compiler import SourceModule
|
|
import numpy as np
|
|
from ktransformers.operators.linear import KTransformersLinear, KLinearMarlin
|
|
from ktransformers.operators.experts import KTransformersExperts, KExpertsTorch
|
|
from ktransformers.util.custom_loader import GGUFLoader, dequantize_q4_k_gpu, dequantize_q4_k
|
|
import torch
|
|
import KTransformersOps
|
|
torch.set_default_dtype(torch.bfloat16)
|
|
import time
|
|
from transformers import (
|
|
AutoConfig,
|
|
)
|
|
|
|
gguf_config = GGUFLoader("/data/Qwen2-57B-A14B-Instruct-GGUF/q4_k_m")
|
|
model_name = "/data/Qwen2-57B-A14B-Instruct"
|
|
key = "blk.0."
|
|
target = "ffn_up_exps.weight"
|
|
|
|
data = gguf_config.get_mmap_tensor(key + target)
|
|
|
|
_, factors, offsets, qs1, qs2= dequantize_q4_k(data)
|
|
factors_cpu = torch.from_numpy(factors)
|
|
offsets_cpu = torch.from_numpy(offsets)
|
|
qs1_cpu = torch.from_numpy(qs1)
|
|
qs2_cpu = torch.from_numpy(qs2)
|
|
|
|
|
|
_, factors, offsets, qs1, qs2 = dequantize_q4_k_gpu(data)
|
|
|
|
print(torch.allclose(factors.cpu(), factors_cpu))
|
|
print(torch.allclose(offsets.cpu(), offsets_cpu))
|
|
print(torch.allclose(qs1.cpu(), qs1_cpu))
|
|
print(torch.allclose(qs2.cpu(), qs2_cpu)) |