def do_work(file,mintime,wayness,lariat_dict):
retval=(None,None,None,None,None)
res=plot.get_data(file,mintime,wayness,lariat_dict)
if (res is None):
return retval
(ts, ld, tmid,
read_rate, write_rate, stall_rate, clock_rate, avx_rate, sse_rate, inst_rate,
meta_rate, l1_rate, l2_rate, l3_rate, load_rate, read_frac, stall_frac) = res
# return (scipy.stats.tmean(stall_frac),
# scipy.stats.tmean((load_rate - (l1_rate + l2_rate +
# l3_rate))/load_rate))
mean_mem_rate=scipy.stats.tmean(read_rate+write_rate)*64.0
ename=ld.exc.split('/')[-1]
ename=tspl_utils.string_shorten(ld.comp_name(ename,ld.equiv_patterns),8)
if ename=='unknown':
return retval
flag=False
if mean_mem_rate < 75.*1000000000./16.:
flag=True
return (scipy.stats.tmean(stall_frac),
scipy.stats.tmean((load_rate - (l1_rate))/load_rate),
scipy.stats.tmean(clock_rate/inst_rate),ename,
flag)
def do_work(file, mintime, wayness, lariat_dict):
bad_retval = (None, None, None, None, None, None, None)
res = plot.get_data(file, mintime, wayness, lariat_dict)
if (res is None):
return bad_retval
(ts, ld, tmid, read_rate, write_rate, stall_rate, clock_rate, avx_rate,
sse_rate, inst_rate, meta_rate, l1_rate, l2_rate, l3_rate, load_rate,
read_frac, stall_frac) = res
mean_mem_rate = scipy.stats.tmean(read_rate + write_rate) * 64.0
ename = ld.exc.split('/')[-1]
if ename == 'unknown':
return bad_retval
ename = ld.comp_name(ename, ld.equiv_patterns)
(s, m, cpi) = (scipy.stats.tmean(stall_frac),
scipy.stats.tmean((load_rate - (l1_rate)) / load_rate),
scipy.stats.tmean(clock_rate / inst_rate))
return (s, m, cpi, ename, ts.j.id, ts.owner, ts.su)
def do_work(file,mintime,wayness,lariat_dict):
bad_retval=(None,None,None,None,None,None,None)
res=plot.get_data(file,mintime,wayness,lariat_dict)
if (res is None):
return bad_retval
(ts, ld, tmid,
read_rate, write_rate, stall_rate, clock_rate, avx_rate, sse_rate, inst_rate,
meta_rate, l1_rate, l2_rate, l3_rate, load_rate, read_frac, stall_frac) = res
mean_mem_rate=scipy.stats.tmean(read_rate+write_rate)*64.0
ename=ld.exc.split('/')[-1]
if ename=='unknown':
return bad_retval
ename=ld.comp_name(ename,ld.equiv_patterns)
(s, m, cpi) = (scipy.stats.tmean(stall_frac),
scipy.stats.tmean((load_rate - (l1_rate))/load_rate),
scipy.stats.tmean(clock_rate/inst_rate))
return (s,m,cpi,ename,ts.j.id,ts.owner,ts.su)
def index():
response = make_response(get_data().to_csv(), 200)
response.mimetype = "text/plain"
return response
matplotlib.use('Agg')
plt.figure(figsize=(17.5, 4.25), frameon=False)
CI_level = 0.6
dir_List = [
('paper/lbdqn/alien-rews', 5000000),
('paper/lbdqn/stargunner-rews', 20000000),
('paper/lbdqn/zaxxon-rews', 15000000),
]
lines = []
for fig_id, dir_entry in enumerate(dir_List):
dir_id, length = dir_entry
title, final_flag = None, False
length //= 50000
data, data_path = get_data(dir_id, final_flag, args.smooth)
for task in data.keys():
plt.style.use('seaborn-whitegrid')
plt.rc('font', family='Times New Roman')
plt.subplot(1, 3, fig_id + 1)
# matplotlib.rcParams['text.usetex'] = True
ax = plt.gca()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)
ax.spines['left'].set_color('black')
ax.spines['bottom'].set_color('black')
if not (title is None):
plt.title(title, size=24.0)
else:
plt.title(task, size=24.0)
def exp4_plot():
datadir = Path("data", "exp4")
zlearner = plot.get_data(Path(datadir, "data00"))
qlearner = plot.get_data(Path(datadir, "data01"))
zlearner_na = plot.get_data(Path(datadir, "data02"))
qlearner_na = plot.get_data(Path(datadir, "data03"))
data = [zlearner, qlearner]
data_na = [zlearner_na, qlearner_na]
basestyle = dict(c="k", lw=0.7)
refstyle = dict(basestyle, c="r", ls="--")
zstyle = dict(basestyle, c="y", ls="-")
qstyle = dict(basestyle, c="b", ls="-.")
zlearner.style.update(zstyle)
qlearner.style.update(qstyle)
zlearner_na.style.update(zstyle)
qlearner_na.style.update(qstyle)
# Figure common setup
cfg = linear.cfg
linear.load_config()
t_range = (0, cfg.final_time)
# All in inches
subsize = (4.05, 0.946)
width = 4.94
top = 0.2
bottom = 0.671765
left = 0.5487688
hspace = 0.2716
# =================
# States and inputs
# =================
figsize, pos = plot.posing(5, subsize, width, top, bottom, left, hspace)
plt.figure(figsize=figsize)
ax = plot.subplot(pos, 0)
[plot.vector_by_index(d, "x", 0)[0] for d in data]
plt.ylabel(r"$x_1$")
# plt.ylim(-2, 2)
plt.legend()
plot.subplot(pos, 1, sharex=ax)
[plot.vector_by_index(d, "x", 1) for d in data]
plt.ylabel(r"$x_2$")
# plt.ylim(-2, 2)
plot.subplot(pos, 2, sharex=ax)
[plot.vector_by_index(d, "u", 0) for d in data]
plt.ylabel(r'$u$')
# plt.ylim(-80, 80)
# ====================
# Parameter estimation
# ====================
ax = plot.subplot(pos, 3)
plot.all(qlearner, "K", style=dict(refstyle, label="True"))
for d in data:
plot.all(
d, "K", is_agent=True,
style=dict(marker="o", markersize=2)
)
plt.ylabel(r"$\hat{K}$")
plt.legend()
# plt.ylim(-70, 30)
plot.subplot(pos, 4, sharex=ax)
plot.all(qlearner, "P", style=dict(qlearner.style, c="r", ls="--"))
for d in data:
plot.all(
d, "P", is_agent=True,
style=dict(marker="o", markersize=2)
)
plt.ylabel(r"$\hat{P}$")
# plt.ylim(-70, 30)
plt.xlabel("Time, sec")
plt.xlim(t_range)
for ax in plt.gcf().get_axes():
ax.label_outer()
# ==================================
# States and inputs (Non-Admissible)
# ==================================
figsize, pos = plot.posing(5, subsize, width, top, bottom, left, hspace)
plt.figure(figsize=figsize)
ax = plot.subplot(pos, 0)
[plot.vector_by_index(d, "x", 0)[0] for d in data_na]
plt.ylabel(r"$x_1$")
# plt.ylim(-2, 2)
plt.legend()
plot.subplot(pos, 1, sharex=ax)
[plot.vector_by_index(d, "x", 1) for d in data_na]
plt.ylabel(r"$x_2$")
# plt.ylim(-2, 2)
plot.subplot(pos, 2, sharex=ax)
[plot.vector_by_index(d, "u", 0) for d in data_na]
plt.ylabel(r'$u$')
# plt.ylim(-80, 80)
# =====================================
# Parameter estimation (Non-Admissible)
# =====================================
ax = plot.subplot(pos, 3)
plot.all(qlearner_na, "K", style=dict(refstyle, label="True"))
for d in data_na:
plot.all(
d, "K", is_agent=True,
style=dict(marker="o", markersize=2)
)
plt.ylabel(r"$\hat{K}$")
plt.legend()
# plt.ylim(-70, 30)
plot.subplot(pos, 4, sharex=ax)
plot.all(qlearner_na, "P", style=refstyle)
for d in data_na:
plot.all(
d, "P", is_agent=True,
style=dict(marker="o", markersize=2)
)
plt.ylabel(r"$\hat{P}$")
# plt.ylim(-70, 30)
plt.xlabel("Time, sec")
plt.xlim(t_range)
for ax in plt.gcf().get_axes():
ax.label_outer()
imgdir = Path("img", datadir.relative_to("data"))
imgdir.mkdir(exist_ok=True)
plt.figure(1)
plt.savefig(Path(imgdir, "figure_1.pdf"), bbox_inches="tight")
plt.figure(2)
plt.savefig(Path(imgdir, "figure_2.pdf"), bbox_inches="tight")
plt.show()
def comp(sqldir, klmdir):
# Data 001 ~ Data 002
sql = plot.get_data(Path(datadir, sqldir))
klm = plot.get_data(Path(datadir, klmdir))
data = [sql, klm]
# data_na = []
basestyle = dict(c="k", lw=0.7)
refstyle = dict(basestyle, c="r", ls="--")
klm_style = dict(basestyle, c="y", ls="-")
sql_style = dict(basestyle, c="b", ls="-.")
klm.style.update(klm_style)
sql.style.update(sql_style)
# zlearner_na.style.update(klm_style)
# qlearner_na.style.update(sql_style)
# Figure common setup
t_range = (0, sql.info["cfg"].env_kwargs.max_t)
# All in inches
subsize = (4.05, 0.946)
width = 4.94
top = 0.2
bottom = 0.671765
left = 0.5487688
hspace = 0.2716
# =================
# States and inputs
# =================
figsize, pos = plot.posing(3, subsize, width, top, bottom, left,
hspace)
plt.figure(figsize=figsize)
ax = plot.subplot(pos, 0)
[plot.vector_by_index(d, "x", 0)[0] for d in data]
plt.ylabel(r"$x_1$")
# plt.ylim(-2, 2)
plt.legend()
plot.subplot(pos, 1, sharex=ax)
[plot.vector_by_index(d, "x", 1) for d in data]
plt.ylabel(r"$x_2$")
# plt.ylim(-2, 2)
plot.subplot(pos, 2, sharex=ax)
[plot.vector_by_index(d, "x", 2) for d in data]
plt.ylabel(r'$x_3$')
# plt.ylim(-80, 80)
plt.xlabel("Time, sec")
plt.xlim(t_range)
for ax in plt.gcf().get_axes():
ax.label_outer()
# ====================
# Parameter estimation
# ====================
figsize, pos = plot.posing(3, subsize, width, top, bottom, left,
hspace)
plt.figure(figsize=figsize)
ax = plot.subplot(pos, 0)
[plot.vector_by_index(d, "u", 0) for d in data]
plt.ylabel(r'$\delta_t$')
plot.subplot(pos, 1, sharex=ax)
plot.all(sql, "K", style=dict(refstyle, label="True"))
for d in data:
plot.all(d,
"K",
is_agent=True,
style=dict(marker="o", markersize=2))
plt.ylabel(r"$\hat{K}$")
plt.legend()
# plt.ylim(-70, 30)
plot.subplot(pos, 2, sharex=ax)
plot.all(sql, "P", style=dict(sql.style, c="r", ls="--"))
for d in data:
plot.all(d,
"P",
is_agent=True,
style=dict(marker="o", markersize=2))
plt.ylabel(r"$\hat{P}$")
# plt.ylim(-70, 30)
plt.xlabel("Time, sec")
plt.xlim(t_range)
for ax in plt.gcf().get_axes():
ax.label_outer()
matplotlib.use('Agg')
fig_cnt = len(experiments_set[args.id])
if fig_cnt==8: fig_w, fig_h = 4, 2
elif fig_cnt==4:
if args.id=='7_slide': fig_w, fig_h = 2, 2
else: fig_w, fig_h = 4, 1
else: fig_w, fig_h = (fig_cnt-1)%3+1, (fig_cnt-1)//3+1
size_w, size_h = size_dict[(fig_w,fig_h)]
plt.figure(figsize=(size_w,size_h),frameon=False)
CI_level = args.CI
legend_id, legend_cnt, lines = 0, 0, []
for fig_id, experiment in enumerate(experiments_set[args.id]):
alg_cnt = 0
data, _ = get_data(final_id[experiment][0])
if len(list(data.keys()))==0: continue
task = list(data.keys())[0]
for alg in list(final_colors[args.id].keys()):
if alg in data[task].keys():
alg_cnt += 1
if alg_cnt>legend_cnt:
legend_cnt = alg_cnt
legend_id = fig_id
for fig_id, experiment in enumerate(experiments_set[args.id]):
if args.id[0]=='9': lines = []
data, _ = get_data(final_id[experiment][0])
plt.subplot(fig_h, fig_w, fig_id+1)
if args.id in ['6_2']:
plt.title(final_id[experiment][1],size=16.0)
# Get Video
vidObj = cv2.VideoCapture(
'/Users/rachelzhou/Research/build_opencv/opencv/samples/python/single/bee6.avi'
)
vidObj.set(1, 50)
ret, frame = vidObj.read()
sec = []
for x in plot.x:
x = x / 23.39
sec.append(x)
t = np.array(sec)
x = np.array(plot.get_data(5))
y = np.array(plot.get_data(0))
z = np.array(plot.get_data(3))
# Configure subplots
fig, axs = plt.subplots(3, 2, sharex='col')
# Initialize the line plots
axs[0, 0].plot(t, x)
axs[0, 0].set_ylabel('velocity' + '\n' + '(pixels per second)')
axs[1, 0].plot(t, y)
axs[1, 0].set_ylabel('angle between' + '\n' + ' antennae (degrees)')
axs[2, 0].plot(t, z)
axs[2, 0].set_ylabel('airflow (meters' + '\n' + ' per second)')