Skip to content

Commit

Permalink
feat: add llm list board
Browse files Browse the repository at this point in the history
  • Loading branch information
@Jcharis
Jcharis committed Jul 25, 2024
1 parent 6b79df7 commit 5aa656b
Show file tree
Hide file tree
Showing 2 changed files with 80 additions and 53 deletions.
17 changes: 17 additions & 0 deletions Dockerfile
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
FROM python:3.9-slim

WORKDIR /app

COPY requirements.txt /app

RUN pip3 install --upgrade pip

RUN pip3 install -r requirements.txt

COPY . /app

EXPOSE 8501

HEALTHCHECK CMD curl --fail http://localhost:8501/_stcore/health

ENTRYPOINT ["streamlit", "run", "streamlit_app.py", "--server.port=8501", "--server.address=0.0.0.0"]
116 changes: 63 additions & 53 deletions app.py
View file
Original file line number Diff line number Diff line change
@@ -1,29 +1,29 @@
import streamlit as st
import streamlit as st
import pandas as pd
import plotly_express as px
import time


def load_data(data):
return pd.read_csv(data)

def stream_text(text:str,delay:float=0.02):

def stream_text(text: str, delay: float = 0.02):
for word in text.split():
yield word + ' '
yield word + " "
time.sleep(delay)

def stream_df(df,delay:float=0.02):

def stream_df(df, delay: float = 0.02):
for i in range(len(df)):
yield df.iloc[[i]]
time.sleep(delay)




df = load_data("data/llm_clean_list.csv")
df_leaderboard = load_data("data/llm_clean_leaderboard.csv")



def about_page():
st.title("About")

Expand All @@ -42,7 +42,7 @@ def ask_llm_page():
model_name = i.get("Model Name")
maintainer = i.get("Maintainer")
size = i.get("Size")
score= i.get("Score")
score = i.get("Score")
context_length = i.get("Context Length")
summary_df = pd.DataFrame([i])
template_text = f"""
Expand All @@ -53,24 +53,24 @@ def ask_llm_page():
Below is a summary table of the model
"""
with st.chat_message("user"):
result = template_text.format(model_name=model_name,maintainer=maintainer)
result = template_text.format(
model_name=model_name, maintainer=maintainer
)
st.write_stream(stream_text(result))
st.write_stream(stream_df(summary_df))



def llm_list_page():
st.title("LLM List")
quick_filters = st.multiselect("Filters",df.columns.tolist())
quick_filters = st.multiselect("Filters", df.columns.tolist())
st.dataframe(df[quick_filters])



def llm_stats_page():
st.title("LLM stats")
metric = st.selectbox("Metric",["Downloads","Likes","Context Length"])
metric = st.selectbox("Metric", ["Downloads", "Likes", "Context Length"])

col1, col2, col3 = st.columns([1,1,1])
col1, col2, col3 = st.columns([1, 1, 1])

with col1:
if st.button("LLM Stats"):
Expand All @@ -84,40 +84,48 @@ def llm_stats_page():
if st.button("LLM Leaderboard"):
show_llm_stats_as_table()


# Select the top 10 models by metric
top_models = df.nlargest(columns=metric,n=10)[["Model Name","Maintainer",metric]]
top_models = df.nlargest(columns=metric, n=10)[["Model Name", "Maintainer", metric]]

# Bar Chart
bar_fig = px.bar(top_models,x="Model Name",y=metric,color="Maintainer")
# Create a Streamlit app
bar_fig = px.bar(top_models, x="Model Name", y=metric, color="Maintainer")
# Create a Streamlit app
st.title(f"Top 10 Models by {metric} - Bar Chart")
st.plotly_chart(bar_fig)

# Pie Chart
pie_fig = px.pie(top_models,names="Model Name",values=metric,color="Maintainer")
# Create a Streamlit app
pie_fig = px.pie(top_models, names="Model Name", values=metric, color="Maintainer")
# Create a Streamlit app
st.title(f"Top 10 Models by {metric} - Pie Chart")
st.plotly_chart(pie_fig)


# Scatter Chart
scatter_fig = px.scatter(top_models,x="Model Name",y=metric,color="Maintainer")
# Create a Streamlit app
scatter_fig = px.scatter(top_models, x="Model Name", y=metric, color="Maintainer")
# Create a Streamlit app
st.title(f"Top 10 Models by {metric} - Scatter Plot")
st.plotly_chart(scatter_fig)


def llm_leaderboard_page():
st.title("LLM Leaderboard")

quick_filters = st.multiselect("Filters ",['Model Name', 'Maintainer', 'License', 'Context Length','Mt Bench', 'Humaneval','Input Priceusd/1M Tokens'],
default= ['Model Name', 'Maintainer', 'License', 'Context Length','Humaneval'
])

quick_filters = st.multiselect(
"Filters ",
[
"Model Name",
"Maintainer",
"License",
"Context Length",
"Mt Bench",
"Humaneval",
"Input Priceusd/1M Tokens",
],
default=["Model Name", "Maintainer", "License", "Context Length", "Humaneval"],
)

st.dataframe(df_leaderboard[quick_filters])

metric = st.selectbox("Metric",["Context Length","Humaneval","MT Bench"])
metric = st.selectbox("Metric", ["Context Length", "Humaneval", "MT Bench"])

col1, col2 = st.columns([1, 1])
with col1:
Expand All @@ -126,36 +134,35 @@ def llm_leaderboard_page():

# Top N Models

# Select the top 10 models by metric
top_models = df_leaderboard.nlargest(columns=metric,n=10)[["Model Name","Maintainer",metric]]
# Select the top 10 models by metric
top_models = df_leaderboard.nlargest(columns=metric, n=10)[
["Model Name", "Maintainer", metric]
]

# Bar Chart
bar_fig = px.bar(top_models,x="Model Name",y=metric,color="Maintainer")
# Create a Streamlit app
bar_fig = px.bar(top_models, x="Model Name", y=metric, color="Maintainer")
# Create a Streamlit app
st.title(f"Top 10 Models by {metric} - Bar Chart")
st.plotly_chart(bar_fig)

# Pie Chart
pie_fig = px.pie(top_models,names="Model Name",values=metric,color="Maintainer")
# Create a Streamlit app
pie_fig = px.pie(top_models, names="Model Name", values=metric, color="Maintainer")
# Create a Streamlit app
st.title(f"Top 10 Models by {metric} - Pie Chart")
st.plotly_chart(pie_fig)


# Scatter Chart
scatter_fig = px.scatter(top_models,x="Model Name",y=metric,color="Maintainer")
# Create a Streamlit app
scatter_fig = px.scatter(top_models, x="Model Name", y=metric, color="Maintainer")
# Create a Streamlit app
st.title(f"Top 10 Models by {metric} - Scatter Plot")
st.plotly_chart(scatter_fig)




@st.experimental_dialog("LLM List - Stats")
def show_llm_stats_as_table():
metric = st.selectbox("Metric ",["Downloads","Likes","Context Length"])
metric = st.selectbox("Metric ", ["Downloads", "Likes", "Context Length"])
# Select the top 10 models by metric
top_models = df.nlargest(columns=metric,n=10)[["Model Name","Maintainer",metric]]
top_models = df.nlargest(columns=metric, n=10)[["Model Name", "Maintainer", metric]]

st.dataframe(top_models)

Expand All @@ -164,28 +171,31 @@ def show_llm_stats_as_table():
def show_llm_list_as_table():
st.dataframe(df)


@st.experimental_dialog("LLM Leaderboard - Stats")
def show_stats_as_table_for_leaderboard():
metric = st.selectbox("Metric ",["Context Length","Humaneval","MT Bench"])
metric = st.selectbox("Metric ", ["Context Length", "Humaneval", "MT Bench"])
# Select the top 10 models by metric
top_models = df_leaderboard.nlargest(columns=metric,n=10)[["Model Name","Maintainer",metric]]
top_models = df_leaderboard.nlargest(columns=metric, n=10)[
["Model Name", "Maintainer", metric]
]

st.dataframe(top_models)


about = st.Page(about_page,title="About",icon=":material/info:")
ask_llm = st.Page(ask_llm_page,title="Ask LLM",icon=":material/chat:")
llm_stats = st.Page(llm_stats_page,title="LLM Stats",icon=":material/list:")
llm_list = st.Page(llm_list_page,title="LLM List",icon=":material/analytics:")
llm_leaderboard = st.Page(llm_leaderboard_page,title="Leaderboard",icon=":material/favorite:")
about = st.Page(about_page, title="About", icon=":material/info:")
ask_llm = st.Page(ask_llm_page, title="Ask LLM", icon=":material/chat:")
llm_stats = st.Page(llm_stats_page, title="LLM Stats", icon=":material/list:")
llm_list = st.Page(llm_list_page, title="LLM List", icon=":material/analytics:")
llm_leaderboard = st.Page(
llm_leaderboard_page, title="Leaderboard", icon=":material/favorite:"
)


# Navigations
pg = st.navigation({"Home":[llm_list,ask_llm,llm_stats,llm_leaderboard],
"About":[about]})
pg = st.navigation(
{"Home": [llm_list, ask_llm, llm_stats, llm_leaderboard], "About": [about]}
)


pg.run()



0 comments on commit 5aa656b

Please sign in to comment.