gavindiaz/NexQuant
1
0
Fork 0
mirror of https://github.com/NicolasBohn/NexQuant.git synced 2026-07-27 23:47:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

chore: data science scenario UI updates (#770)

Browse Source 
* fix some pandas warning

* fix load time logic

* optimize load_time logic

* fix CI

* fix bug
This commit is contained in:
XianBW
2025-04-08 19:33:20 +08:00
committed by GitHub
parent f092739ab5
commit 6c8fdeaa41
2 changed files with 56 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rdagent/log/ui/ds_summary.py
+44 -19
View File
@@ -14,7 +14,7 @@ from rdagent.log.ui.conf import UI_SETTING
from rdagent.log.ui.ds_trace import load_times
def get_exec_time(stdout_p: Path):
def get_script_time(stdout_p: Path):
with stdout_p.open("r") as f:
first_line = next(f).strip()
last_line = deque(f, maxlen=1).pop().strip()
@@ -55,20 +55,23 @@ def get_summary_df(log_folders: list[str]) -> tuple[dict, pd.DataFrame]:
for k, v in s.items():
stdout_p = Path(lf) / f"{k}.stdout"
if stdout_p.exists():
v["exec_time"] = get_exec_time(stdout_p)
v["script_time"] = get_script_time(stdout_p)
else:
v["exec_time"] = None
v["script_time"] = None
exp_gen_time = timedelta()
coding_time = timedelta()
running_time = timedelta()
if state.show_times_info:
times_info = load_times(Path(lf) / k)
for time_info in times_info.values():
exp_gen_time += time_info[0].end - time_info[0].start
all_time = timedelta()
times_info = load_times(Path(lf) / k)
for time_info in times_info.values():
all_time += sum((ti.end - ti.start for ti in time_info), timedelta())
exp_gen_time += time_info[0].end - time_info[0].start
if len(time_info) > 1:
coding_time += time_info[1].end - time_info[1].start
if len(time_info) > 2:
running_time += time_info[2].end - time_info[2].start
if len(time_info) > 2:
running_time += time_info[2].end - time_info[2].start
v["exec_time"] = str(all_time).split(".")[0]
v["exp_gen_time"] = str(exp_gen_time).split(".")[0]
v["coding_time"] = str(coding_time).split(".")[0]
v["running_time"] = str(running_time).split(".")[0]
@@ -84,6 +87,7 @@ def get_summary_df(log_folders: list[str]) -> tuple[dict, pd.DataFrame]:
base_df = pd.DataFrame(
columns=[
"Competition",
"Script Time",
"Exec Time",
"Exp Gen",
"Coding",
@@ -132,6 +136,7 @@ def get_summary_df(log_folders: list[str]) -> tuple[dict, pd.DataFrame]:
for k, v in summary.items():
loop_num = v["loop_num"]
base_df.loc[k, "Competition"] = v["competition"]
base_df.loc[k, "Script Time"] = v["script_time"]
base_df.loc[k, "Exec Time"] = v["exec_time"]
base_df.loc[k, "Exp Gen"] = v["exp_gen_time"]
base_df.loc[k, "Coding"] = v["coding_time"]
@@ -205,6 +210,8 @@ def get_summary_df(log_folders: list[str]) -> tuple[dict, pd.DataFrame]:
def num2percent(num: int, total: int, show_origin=True) -> str:
num = int(num)
total = int(total)
if show_origin:
return f"{num} ({round(num / total * 100, 2)}%)"
return f"{round(num / total * 100, 2)}%"
@@ -212,6 +219,20 @@ def num2percent(num: int, total: int, show_origin=True) -> str:
def percent_df(df: pd.DataFrame, show_origin=True) -> pd.DataFrame:
base_df = df.copy(deep=True)
# Convert columns to object dtype so we can store strings like "14 (53.85%)" without warnings
columns_to_convert = [
"Successful Final Decision",
"Made Submission",
"Valid Submission",
"Above Median",
"Bronze",
"Silver",
"Gold",
"Any Medal",
]
base_df[columns_to_convert] = base_df[columns_to_convert].astype(object)
for k in base_df.index:
loop_num = int(base_df.loc[k, "Total Loops"])
if loop_num != 0:
@@ -231,6 +252,7 @@ def percent_df(df: pd.DataFrame, show_origin=True) -> pd.DataFrame:
base_df.loc[k, "Silver"] = num2percent(base_df.loc[k, "Silver"], loop_num, show_origin)
base_df.loc[k, "Gold"] = num2percent(base_df.loc[k, "Gold"], loop_num, show_origin)
base_df.loc[k, "Any Medal"] = num2percent(base_df.loc[k, "Any Medal"], loop_num, show_origin)
return base_df
@@ -275,12 +297,13 @@ def all_summarize_win():
base_df = percent_df(base_df)
base_df.insert(0, "Select", True)
base_df = st.data_editor(
base_df.style.applymap(
lambda x: "background-color: #F0F8FF",
base_df.style.apply(
lambda col: col.map(lambda val: "background-color: #F0F8FF"),
subset=["Baseline Score", "Bronze Threshold", "Silver Threshold", "Gold Threshold", "Medium Threshold"],
axis=0,
)
.applymap(
lambda x: "background-color: #FFFFE0",
.apply(
lambda col: col.map(lambda val: "background-color: #FFFFE0"),
subset=[
"Ours - Base",
"Ours vs Base",
@@ -288,23 +311,27 @@ def all_summarize_win():
"Ours vs Silver",
"Ours vs Gold",
],
axis=0,
)
.applymap(
lambda x: "background-color: #E6E6FA",
.apply(
lambda col: col.map(lambda val: "background-color: #E6E6FA"),
subset=[
"Script Time",
"Exec Time",
"Exp Gen",
"Coding",
"Running",
],
axis=0,
)
.applymap(
lambda x: "background-color: #F0FFF0",
.apply(
lambda col: col.map(lambda val: "background-color: #F0FFF0"),
subset=[
"Best Result",
"SOTA Exp",
"SOTA Exp Score",
],
axis=0,
),
column_config={
"Select": st.column_config.CheckboxColumn("Select", default=True, help="Stat this trace.", disabled=False),
@@ -408,8 +435,6 @@ def all_summarize_win():
st.json(v["valid_scores"])
with st.sidebar:
st.toggle("Show Times Info (Slowly)", key="show_times_info")
with st.container(border=True):
if st.toggle("近3天平均", key="show_3days"):
days_summarize_win()
rdagent/log/ui/ds_trace.py
+12 -17
View File
@@ -50,22 +50,16 @@ def convert_defaultdict_to_dict(d):
@st.cache_data(persist=True)
def load_times(log_path: Path):
"""加载时间数据"""
times = defaultdict(lambda: defaultdict(dict))
for msg in FileStorage(log_path).iter_msg():
if msg.tag and "llm" not in msg.tag and "session" not in msg.tag:
li, fn = extract_loopid_func_name(msg.tag)
if li:
li = int(li)
try:
session_path = log_path / "__session__"
max_li = max(int(p.name) for p in session_path.iterdir() if p.is_dir() and p.name.isdigit())
max_step = max(int(p.name.split("_")[0]) for p in (session_path / str(max_li)).iterdir() if p.is_file())
rdloop_obj_p = next((session_path / str(max_li)).glob(f"{max_step}_*"))
# read times
loop_obj_path = log_path / "__session__" / f"{li}" / "4_record"
if loop_obj_path.exists():
try:
times[li] = DataScienceRDLoop.load(loop_obj_path, do_truncate=False).loop_trace[li]
except Exception as e:
pass
return convert_defaultdict_to_dict(times)
rd_times = DataScienceRDLoop.load(rdloop_obj_p, do_truncate=False).loop_trace
except:
rd_times = {}
return rd_times
@st.cache_data(persist=True)
@@ -436,8 +430,9 @@ def summarize_data():
df.loc[loop, "Time"] = str(sum((i.end - i.start for i in state.times[loop]), timedelta())).split(".")[0]
exp_gen_time = state.times[loop][0].end - state.times[loop][0].start
df.loc[loop, "Exp Gen"] = str(exp_gen_time).split(".")[0]
coding_time = state.times[loop][1].end - state.times[loop][1].start
df.loc[loop, "Coding"] = str(coding_time).split(".")[0]
if len(state.times[loop]) > 1:
coding_time = state.times[loop][1].end - state.times[loop][1].start
df.loc[loop, "Coding"] = str(coding_time).split(".")[0]
if len(state.times[loop]) > 2:
running_time = state.times[loop][2].end - state.times[loop][2].start
df.loc[loop, "Running"] = str(running_time).split(".")[0]