pratikgx's picture
Add FastFrames-S M5 Pro Mac mini bench (MLX port data and notes, no weights)
c5627a4 verified
Raw History Blame Contribute Delete
1.61 kB
import re, matplotlib; matplotlib.use("Agg"); import matplotlib.pyplot as plt
t=[float(m) for m in re.findall(r"latent frame \d+\s+([\d.]+) ms",open("clip.log").read())]
fig=plt.figure(figsize=(16,9),dpi=100,facecolor="#000")
fig.text(0.5,0.92,"Apple FastFrames-S on a 64GB M5 Pro Mac mini (MLX)",ha="center",color="#eee",fontsize=30,weight="bold")
fig.text(0.5,0.865,"832x480 · 81 frames (5 s at 16 fps) · seed 0 · one run",ha="center",color="#888",fontsize=16)
ax=fig.add_axes([0.07,0.2,0.55,0.56],facecolor="#000")
ax.bar(range(len(t)),[x/1000 for x in t],color="#e0a040",width=0.7)
ax.set_title("Seconds per latent frame (each = ~4 video frames)",color="#ccc",fontsize=18,pad=12)
ax.tick_params(colors="#bbb",labelsize=13); ax.set_xlabel("latent frame",color="#999",fontsize=14)
for s in ax.spines.values(): s.set_visible(False)
ax.grid(axis="y",color="#222")
ax.text(0,t[0]/1000+0.05,"%.1fs\nwarm-up"%(t[0]/1000),ha="center",color="#eee",fontsize=12)
ax.text(len(t)-1,t[-1]/1000+0.05,"%.1fs"%(t[-1]/1000),ha="center",color="#eee",fontsize=12)
stats=[("30.4 s","total generation"),("376 ms","per output frame"),("22.9 GB","peak MLX memory"),("~6x","off real time")]
for i,(v,l) in enumerate(stats):
y=0.70-i*0.14
fig.text(0.80,y,v,ha="center",color="#eee",fontsize=34,weight="bold",family="monospace")
fig.text(0.80,y-0.04,l,ha="center",color="#888",fontsize=15)
fig.text(0.5,0.06,"Frames are generated one after another, and each one takes a little longer than the last",ha="center",color="#aaa",fontsize=16)
fig.savefig("fastframes_mac_chart.png",facecolor="#000")
print(len(t), sum(t)/1000)