vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Augmented benchmark stats (#929)

Browse source 
* Augmented benchmark stats v1

* output instead of coherence

* populate bench flags as a flags field instead of multiple lines

---------

Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com>
This commit is contained in:
Nexesenex 2024-06-18 15:30:36 +02:00 • committed by GitHub
parent a0ecd0d8e6
commit 153527745b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 7 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

12
koboldcpp.py
View file

@ -3577,26 +3577,28 @@ def main(launch_args,start_server=True):
s_pp = float(benchmaxctx-benchlen)/t_pp
s_gen = float(benchlen)/t_gen
datetimestamp = datetime.now(timezone.utc)
benchflagstr = f"NoAVX2={args.noavx2} Threads={args.threads} HighPriority={args.highpriority} NoBlas={args.noblas} Cublas_Args={args.usecublas} Tensor_Split={args.tensor_split} BlasThreads={args.blasthreads} BlasBatchSize={args.blasbatchsize} FlashAttention={args.flashattention} KvCache={args.quantkv}"
print(f"\nBenchmark Completed - v{KcppVersion} Results:\n======")
print(f"Flags: {benchflagstr}")
print(f"Timestamp: {datetimestamp}")
print(f"Backend: {libname}")
print(f"Layers: {args.gpulayers}")
print(f"Model: {benchmodel}")
print(f"MaxCtx: {benchmaxctx}")
print(f"GenAmount: {benchlen}\n-----")
print(f"ProcessingTime: {t_pp:.2f}s")
print(f"ProcessingTime: {t_pp:.3f}s")
print(f"ProcessingSpeed: {s_pp:.2f}T/s")
print(f"GenerationTime: {t_gen:.2f}s")
print(f"GenerationTime: {t_gen:.3f}s")
print(f"GenerationSpeed: {s_gen:.2f}T/s")
print(f"TotalTime: {(t_pp+t_gen):.2f}s")
print(f"TotalTime: {(t_pp+t_gen):.3f}s")
print(f"Output: {result}\n-----")
if save_to_file:
try:
with open(args.benchmark, "a") as file:
file.seek(0, 2)
if file.tell() == 0: #empty file
file.write(f"Timestamp,Backend,Layers,Model,MaxCtx,GenAmount,ProcessingTime,ProcessingSpeed,GenerationTime,GenerationSpeed,TotalTime,Output")
file.write(f"\n{datetimestamp},{libname},{args.gpulayers},{benchmodel},{benchmaxctx},{benchlen},{t_pp:.2f},{s_pp:.2f},{t_gen:.2f},{s_gen:.2f},{(t_pp+t_gen):.2f},{result}")
file.write(f"Timestamp,Backend,Layers,Model,MaxCtx,GenAmount,ProcessingTime,ProcessingSpeed,GenerationTime,GenerationSpeed,TotalTime,Output,Flags")
file.write(f"\n{datetimestamp},{libname},{args.gpulayers},{benchmodel},{benchmaxctx},{benchlen},{t_pp:.2f},{s_pp:.2f},{t_gen:.2f},{s_gen:.2f},{(t_pp+t_gen):.2f},{result},{benchflagstr}")
except Exception as e:
print(f"Error writing benchmark to file: {e}")
global using_gui_launcher