add one picture

.gitignore

@@ -1,2 +1,4 @@
 get_loss/__pycache__/
-*.pyc
+*.pyc
+gradio_cached_examples/
+get_loss/__pycache__/get_loss_hf.cpython-310.pyc

app.py

@@ -4,6 +4,8 @@ from transformers import AutoTokenizer
 from get_loss.get_loss_hf import run_get_loss
 import pdb
 from types import SimpleNamespace
+import pandas as pd
+import plotly.express as px
 # os.system('git clone https://github.com/EleutherAI/lm-evaluation-harness')
 # os.system('cd lm-evaluation-harness')
 # os.system('pip install -e .')
@@ -56,6 +58,45 @@ def get_text(ids_list=[0.1,0.7], tokenizer=None):
 #     # 这里只是简单地返回 IDs 和损失值，但是可以根据实际需求添加颜色或其他样式
 #     return [1, 2], [0.1, 0.7]
 
+def harness_eval(question, choices, answer_index, model=None,tokenizer=None):
+    '''
+    use harness to test one question,  can specify the model, (extract or ppl)
+    '''
+    # TODO add the model and its score
+    # torch.nn.functional.softmax(output.logits, dim=0)
+    # topk = torch.topk(output.logits, 5)
+    
+    return {'A':0.5, 'B':0.3, 'C':0.1, 'D':0.1}
+
+
+
+
+def plotly_plot():#(df, x, y, color,title, x_title, y_title):
+        # plotly_plot(sample_df, 'date', 'loss_mean_at_1000', 'model','ppl with time', 'time', 'ppl')
+    df=pd.read_csv('./data/tmp.csv')
+    df['date'] = pd.to_datetime(df['date'])
+    # sort by date
+    df.sort_values(by='date', inplace=True)
+
+    # use a dic to filter the dataframe
+    df = df[df['file_name'] == 'arxiv_computer_science']
+
+    x,y,color,title, x_title, y_title='date', 'loss_mean_at_1000', 'model','ppl with time', 'time', 'ppl'
+
+    fig = px.line(df, x=x, y=y, color=color,title=title)
+    fig.update_xaxes(title_text=x_title)
+    fig.update_yaxes(title_text=y_title)
+    # fig.update_layout()
+    return fig
+
+# def plotly_plot(df, x, y, color, title, x_title, y_title):
+#     fig = px.line(df, x=x, y=y, color=color, title=title)
+#     fig.update_xaxes(title_text=x_title)
+#     fig.update_yaxes(title_text=y_title)
+#     return fig
+
+
+
 
 def color_pipeline(texts=["Hi","FreshEval","!"],  model=None):
     """
@@ -83,17 +124,19 @@ with gr.Blocks() as demo:
     with gr.Tab("color your text"):
         with gr.Row():
             text_input = gr.Textbox(label="input text", placeholder="input your text here...")
+            # file_input = gr.File(file_count="multiple",label='to add content')#
             # TODO craw and drop the file
 
             # loss_input = gr.Number(label="loss")
-            model_input = gr.Textbox(label="model name", placeholder="input your model name here... now I am trying phi-2...")
-            output_box=gr.HighlightedText(label="colored text")
-            # gr.Examples(
-            #     [
-            #         # ["Hi FreshEval !", "microsoft/phi-2"],
-            #         ["Hello FreshBench !", "/home/sribd/chenghao/models/phi-2"],
-            #     ],
-            #     [text_input, model_input],
+            model_input = gr.Textbox(label="model name", placeholder="input your model name here... now I am trying phi-2...")#TODO make a choice here
+        output_box=gr.HighlightedText(label="colored text")#,interactive=True
+        
+        gr.Examples(
+                [
+                    ["Hi FreshEval !", "microsoft/phi-2"],
+                    ["Hello FreshBench !", "/home/sribd/chenghao/models/phi-2"],
+                ],
+                [text_input, model_input],)
             #     cache_examples=True,
             #     # cache_examples=False,
             #     fn=color_pipeline,
@@ -103,10 +146,15 @@ with gr.Blocks() as demo:
             # TODO maybe add our own models
 
 
-            color_text_output = gr.HTML(label="colored text")
-
+        color_text_output = gr.HTML(label="colored text")
         color_text_button = gr.Button("color the text").click(color_pipeline, inputs=[text_input, model_input], outputs=output_box)
 
+        # markdown
+        gr.Markdown('### How to use this app')
+        
+
+
+
 
         date_time_input = gr.Textbox(label="the date when the text is generated")#TODO add date time input
         description_input = gr.Textbox(label="description of the text")
@@ -117,18 +165,37 @@ with gr.Blocks() as demo:
         '''
         use extract, or use ppl
         '''
-        question=gr.Textbox(placeholder='input your question here...')
-        answer=gr.Textbox(placeholder='input your answer here...')
-        other_choices=gr.Textbox(placeholder='input your other choices here...')
+        question=gr.Textbox(label="input question", placeholder='input your question here...')
+        answer_index=gr.Textbox(label="right answer index", placeholder='index for right anser here, start with 0')#TODO add multiple choices, 
+        choices=gr.Textbox(placeholder='input your other choices here...')
         
-        test_button=gr.Button('test').click()
+        # test_button=gr.Button('test').click(harness_eval())# TODO figure out the input and output
+
+        answer_type=gr.Dropdown(label="answer type", choices=['extract', 'ppl'])
+        #TODO add the model and its score
+        answer_label=gr.Label('the answers\'s detail')# RETURN the answer and its score,in the form of dic{str: float}
+
+        test_question_button=gr.Button('test question').click(harness_eval,inputs=[question, choices, answer_index ,answer_type],outputs=[answer_label])
+
+        forecast_q='A Ukrainian counteroffensive began in 2023, though territorial gains by November 2023 were limited (Economist, BBC, Newsweek). The question will be suspended on 31 July 2024 and the outcome determined using data as reported in the Brookings Institution\'s "Ukraine Index" (Brookings Institution - Ukraine Index, see "Percentage of Ukraine held by Russia" chart). If there is a discrepancy between the chart data and the downloaded data (see "Get the data" within the "NET TERRITORIAL GAINS" chart border), the downloaded data will be used for resolution.'
+        answer_list=['Less than 5%','At least 5%, but less than 10%','At least 10%, but less than 15%','At least 15%, but less than 20%','20% or more' ]
+         
+        gr.Examples([
+                    [forecast_q, '&&&&&&'.join(answer_list), '0']
+                     ],
+                    [question, choices, answer_index])
+
+        date_time_input = gr.Textbox(label="the date when the text is generated")#TODO add date time input
+        description_input = gr.Textbox(label="description of the text")
+        submit_button = gr.Button("submit a post or record").click()
+
         #TODO add the model and its score
 
         def test_question(question, answer, other_choices):
             '''
             use extract, or use ppl
             '''
-            answer_ppl, other_choices_ppl = get_ppl(question, answer, other_choices)
+            answer_ppl, other_choices_ppl =  (question, answer, other_choices)
             return answer_ppl, other_choices_ppl
 
 
@@ -139,6 +206,23 @@ with gr.Blocks() as demo:
         '''
         # load the json file with time,
 
+        # sample_df=pd.DataFrame({'time':pd.date_range('2021-01-01', periods=6), 'ppl': [1,2,3,4,5,6]})
+        pd_df=pd.read_csv('./data/tmp.csv')
+        pd_df['date'] = pd.to_datetime(pd_df['date'])
+        print(pd_df.head)
+        # gr_df=gr.Dataframe(pd_df)
+        gr_df=pd_df
+
+
+        # print(gr_df.head)
+        print('done')
+        # sample
+        plot=gr.Plot(label='model text ppl')
+        # plotly_plot(gr_df, 'date', 'loss_mean_at_1000', 'model','ppl with time', 'time', 'ppl')
+        # draw_pic_button=gr.Button('draw the pic').click(plotly_plot,inputs=['gr_df', 'date', 'loss_mean_at_1000', 'model','ppl with time', 'time', 'ppl'],outputs=[plot])
+        draw_pic_button=gr.Button('draw the pic').click(plotly_plot,inputs=[],outputs=[plot])
+
+
     
     with gr.Tab("model quesion acc with time"):
         '''

data/ob.csv

@@ -0,0 +1,32 @@
+,day,Mexico,UK
+0,60,98458.24943217951,39579.56206409874
+1,61,100786.91721556247,40515.671040143054
+2,62,103131.45192653117,41458.158415648744
+3,63,105491.7031365984,42406.96371936348
+4,64,107867.5242167411,43362.02800739426
+5,65,110258.77218336452,44323.29380128607
+6,66,112665.30755279331,45290.70502952821
+7,67,115086.99420369453,46264.206972249296
+8,68,117523.69924688818,47243.74620888196
+9,69,119975.29290204526,48229.270568596265
+10,70,122441.64838081415,49220.72908331732
+11,71,124922.64177595275,50218.07194315709
+12,72,127418.1519560776,51221.25045410429
+13,73,129928.06046567051,52230.21699782768
+14,74,132452.25143001205,53244.92499345999
+15,75,134990.61146473512,54265.32886123903
+16,76,137543.02958971576,55291.38398789239
+17,77,140109.39714703834,56323.046693659926
+18,78,142689.60772279178,57360.2742008565
+19,79,145283.55707247148,58403.02460388389
+20,80,147891.1430497762,59451.256840607704
+21,81,150512.26553860537,60504.930665020765
+22,82,153146.82638807516,61564.006621119916
+23,83,155794.7293503834,62628.4460179282
+24,84,158455.88002136638,63698.21090559912
+25,85,161130.1857835995,64773.26405254332
+26,86,163817.55575190464,65853.56892352313
+27,87,166517.90072113517,66939.08965866231
+28,88,169231.13311611864,68029.79105332344
+29,89,171957.1669436441,69125.63853880709
+30,90,174695.91774638867,70226.59816383067

data/ob.py

@@ -0,0 +1,15 @@
+import pandas as pd
+import plotly.express as px
+
+def plotly_plot(df, x, y, color, title, x_title, y_title):
+    fig = px.line(df, x=x, y=y, color=color, title=title)
+    fig.update_xaxes(title_text=x_title)
+    fig.update_yaxes(title_text=y_title)
+    return fig
+
+pd_df = pd.read_csv('./tmp.csv')
+pd_df['date'] = pd.to_datetime(pd_df['date'])
+
+fig=plotly_plot(pd_df, 'date', 'loss_mean_at_1000', 'model', 'ppl with time', 'time', 'ppl')
+fig.show()
+

plot.py

@@ -0,0 +1,79 @@
+import altair
+
+import gradio as gr
+from math import sqrt
+import matplotlib.pyplot as plt
+import numpy as np
+import plotly.express as px
+import pandas as pd
+import pdb
+
+def outbreak(plot_type, r, month, countries, social_distancing):
+    months = ["January", "February", "March", "April", "May"]
+    m = months.index(month)
+    start_day = 30 * m
+    final_day = 30 * (m + 1)
+    x = np.arange(start_day, final_day + 1)
+    pop_count = {"USA": 350, "Canada": 40, "Mexico": 300, "UK": 120}
+    if social_distancing:
+        r = sqrt(r)
+    # df = pd.DataFrame({"day": x})
+    # for country in countries:
+    #     df[country] = x ** (r) * (pop_count[country] + 1)
+    df=pd.read_csv('./data/tmp.csv')
+    print(df.head())
+    # pdb.set_trace()
+
+    if plot_type == "Matplotlib":
+        fig = plt.figure()
+        plt.plot(df["day"], df[countries].to_numpy())
+        plt.title("Outbreak in " + month)
+        plt.ylabel("Cases")
+        plt.xlabel("Days since Day 0")
+        plt.legend(countries)
+        return fig
+    elif plot_type == "Plotly":
+        fig = px.line(df, x="day", y=countries)
+        fig.update_layout(
+            title="Outbreak in " + month,
+            xaxis_title="Cases",
+            yaxis_title="Days Since Day 0",
+        )
+        return fig
+    elif plot_type == "Altair":
+        df = df.melt(id_vars="day").rename(columns={"variable": "country"})
+        fig = altair.Chart(df).mark_line().encode(x="day", y='value', color='country')
+        return fig
+    else:
+        raise ValueError("A plot type must be selected")
+
+
+inputs = [
+    gr.Dropdown(["Matplotlib", "Plotly", "Altair"], label="Plot Type"),
+    gr.Slider(1, 4, 3.2, label="R"),
+    gr.Dropdown(["January", "February", "March", "April", "May"], label="Month"),
+    gr.CheckboxGroup(
+        ["USA", "Canada", "Mexico", "UK"], label="Countries", value=["USA", "Canada"]
+    ),
+    gr.Checkbox(label="Social Distancing?"),
+]
+outputs = gr.Plot()
+
+demo = gr.Interface(
+    fn=outbreak,
+    inputs=inputs,
+    outputs=outputs,
+    examples=[
+        ["Matplotlib", 2, "March", ["Mexico", "UK"], True],
+        ["Altair", 2, "March", ["Mexico", "Canada"], True],
+        ["Plotly", 3.6, "February", ["Canada", "Mexico", "UK"], False],
+    ],
+    cache_examples=True,
+)
+
+if __name__ == "__main__":
+    demo.launch()
+
+
+
+