authwork
authwork
Many thanks for your help @ebubae @beat-buesser Then, when I execure to ```graph = tf.get_default_graph()``` It has the following problem (I have tried this before, that's why I said it...
I used the most recent version of both and remove the limitation for 2.2.4 in ART.
> on the server side, it will wait all workers' data to merge them before sending back ACK for workers' requests Exactly, but this is the description of BSP model....
V100 fastllm - - - - - - - - - - - - - - - - -- f16, batch = 1 batch: 1 output 336 tokens (正常) use...
输入长度是同样的Prompt "Building a website can be done in 10 simple steps.” 详情见 #17 我这边的V100是16G显存的型号
@ztxz16 请帮忙看看这个小问题 V100测试LLama: [ user: "Below is an instruction that describes a task. Write a response that appropriately completes the request. ### Instruction: Building a website can be done in...
@ztxz16 在V100上测试也有一点点问题, 同样的Prompt "Building a website can be done in 10 simple steps.": i8, batch=2 batch: 2 output 670 tokens (正常) use 43.564709 s speed = 15.379421 tokens / s...
> 使用楼主的代码测试 我的硬件环境： cuda 11.2 gcc 7.5 v100 CmakeLists.txt 文件第20行修改为：set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pthread --std=c++17 -O2") > > 测试速度如下： fp16 generate token number 2330 time consume 43.830317974090576 s 18.81129526785003 ms/token > >...