Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -19,66 +19,61 @@ import httpx
|
|
19 |
import pandas as pd
|
20 |
import datasets as ds
|
21 |
|
22 |
-
# -------------------------------------------- For Memory - you will need to set up a dataset and HF_TOKEN ---------
|
23 |
-
#UseMemory=False
|
24 |
UseMemory=True
|
25 |
-
|
26 |
-
|
27 |
-
DATASET_REPO_URL="https://huggingface.co/datasets/awacke1/ChatbotMemory.csv"
|
28 |
-
DATASET_REPO_ID="awacke1/ChatbotMemory.csv"
|
29 |
-
DATA_FILENAME="ChatbotMemory.csv"
|
30 |
-
DATA_FILE=os.path.join("data", DATA_FILENAME)
|
31 |
HF_TOKEN=os.environ.get("HF_TOKEN")
|
32 |
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
|
|
|
|
|
|
46 |
|
47 |
-
def get_df(name: str):
|
48 |
-
dataset = load_dataset(str, split="train")
|
49 |
-
return dataset
|
50 |
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
#repo.git_pull() # test repull to avoid out of sync rrepo error due to others commits
|
55 |
-
#repo = repo.git_pull(local_dir="data", clone_from=DATASET_REPO_URL, use_auth_token=HF_TOKEN) # test repull to avoid out of sync rrepo error due to others commits
|
56 |
-
#repo = Repository(local_dir="data", clone_from=DATASET_REPO_URL, use_auth_token=HF_TOKEN)
|
57 |
-
#
|
58 |
-
with open(DATA_FILE, "a") as csvfile:
|
59 |
-
writer = csv.DictWriter(csvfile, fieldnames=[ "time", "message", "name", ])
|
60 |
-
writer.writerow(
|
61 |
-
{"time": str(datetime.now()), "message": message.strip(), "name": name.strip() }
|
62 |
-
)
|
63 |
-
#repo.git_pull(rebase=True)
|
64 |
-
commit_url = repo.push_to_hub()
|
65 |
-
|
66 |
-
# test api retrieval of any dataset that is saved, then return it...
|
67 |
-
# app = FastAPI()
|
68 |
-
# see: https://gradio.app/sharing_your_app/#api-page
|
69 |
-
|
70 |
-
# f=get_df(DATASET_REPO_ID)
|
71 |
-
# print(f)
|
72 |
-
#return commit_url
|
73 |
-
return ""
|
74 |
-
# ----------------------------------------------- For Memory
|
75 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
mname = "facebook/blenderbot-400M-distill"
|
77 |
model = BlenderbotForConditionalGeneration.from_pretrained(mname)
|
78 |
tokenizer = BlenderbotTokenizer.from_pretrained(mname)
|
79 |
|
80 |
def take_last_tokens(inputs, note_history, history):
|
81 |
-
"""Filter the last 128 tokens"""
|
82 |
if inputs['input_ids'].shape[1] > 128:
|
83 |
inputs['input_ids'] = torch.tensor([inputs['input_ids'][0][-128:].tolist()])
|
84 |
inputs['attention_mask'] = torch.tensor([inputs['attention_mask'][0][-128:].tolist()])
|
@@ -87,7 +82,6 @@ def take_last_tokens(inputs, note_history, history):
|
|
87 |
return inputs, note_history, history
|
88 |
|
89 |
def add_note_to_history(note, note_history):# good example of non async since we wait around til we know it went okay.
|
90 |
-
"""Add a note to the historical information"""
|
91 |
note_history.append(note)
|
92 |
note_history = '</s> <s>'.join(note_history)
|
93 |
return [note_history]
|
@@ -96,12 +90,21 @@ title = "💬ChatBack🧠💾"
|
|
96 |
description = """Chatbot With persistent memory dataset allowing multiagent system AI to access a shared dataset as memory pool with stored interactions.
|
97 |
Current Best SOTA Chatbot: https://huggingface.co/facebook/blenderbot-400M-distill?text=Hey+my+name+is+ChatBack%21+Are+you+ready+to+rock%3F """
|
98 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
def chat(message, history):
|
100 |
history = history or []
|
101 |
if history:
|
102 |
history_useful = ['</s> <s>'.join([str(a[0])+'</s> <s>'+str(a[1]) for a in history])]
|
103 |
else:
|
104 |
history_useful = []
|
|
|
105 |
history_useful = add_note_to_history(message, history_useful)
|
106 |
inputs = tokenizer(history_useful, return_tensors="pt")
|
107 |
inputs, history_useful, history = take_last_tokens(inputs, history_useful, history)
|
@@ -114,46 +117,28 @@ def chat(message, history):
|
|
114 |
df=pd.DataFrame()
|
115 |
|
116 |
if UseMemory:
|
117 |
-
|
118 |
-
|
119 |
-
df =
|
120 |
-
|
121 |
-
|
122 |
-
return history, df
|
123 |
-
#return df
|
124 |
-
#return history, df
|
125 |
|
126 |
|
127 |
-
#gr.Interface(
|
128 |
-
# fn=chat,
|
129 |
-
# theme="huggingface",
|
130 |
-
# css=".footer {display:none !important}",
|
131 |
-
# inputs=["text", "state"],
|
132 |
-
# #outputs=["chatbot", "state", "text"],
|
133 |
-
# outputs=["chatbot", "state", "dataframe"],
|
134 |
-
# title=title,
|
135 |
-
# allow_flagging="never",
|
136 |
-
# description=f"Gradio chatbot backed by memory in a dataset repository.",
|
137 |
-
# article=f"The memory dataset for saves is [{DATASET_REPO_URL}]({DATASET_REPO_URL}) And here: https://huggingface.co/spaces/awacke1/DatasetAnalyzer Code and datasets on chat are here hf tk: https://paperswithcode.com/datasets?q=chat&v=lst&o=newest"
|
138 |
-
# ).launch(debug=True)
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
with gr.Blocks() as demo:
|
143 |
-
gr.Markdown("<h1><center>🍰Gradio chatbot backed by
|
144 |
-
#gr.Markdown("The memory dataset for saves is [{DATASET_REPO_URL}]({DATASET_REPO_URL}) And here: https://huggingface.co/spaces/awacke1/DatasetAnalyzer Code and datasets on chat are here hf tk: https://paperswithcode.com/datasets?q=chat&v=lst&o=newest")
|
145 |
|
146 |
with gr.Row():
|
147 |
t1 = gr.Textbox(lines=1, default="", label="Chat Text:")
|
148 |
-
b1 = gr.Button("
|
149 |
|
150 |
with gr.Row(): # inputs and buttons
|
151 |
s1 = gr.State([])
|
152 |
-
s2 = gr.Markdown()
|
153 |
-
with gr.Row():
|
154 |
df1 = gr.Dataframe(wrap=True, max_rows=1000, overflow_row_behaviour= "paginate")
|
155 |
-
|
156 |
-
|
157 |
-
|
158 |
|
159 |
-
|
|
|
|
|
|
19 |
import pandas as pd
|
20 |
import datasets as ds
|
21 |
|
|
|
|
|
22 |
UseMemory=True
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
HF_TOKEN=os.environ.get("HF_TOKEN")
|
24 |
|
25 |
+
def SaveResult(text, outputfileName):
|
26 |
+
basedir = os.path.dirname(__file__)
|
27 |
+
savePath = outputfileName
|
28 |
+
print("Saving: " + text + " to " + savePath)
|
29 |
+
from os.path import exists
|
30 |
+
file_exists = exists(savePath)
|
31 |
+
if file_exists:
|
32 |
+
with open(outputfileName, "a") as f: #append
|
33 |
+
f.write(str(text.replace("\n"," ")))
|
34 |
+
f.write('\n')
|
35 |
+
else:
|
36 |
+
with open(outputfileName, "w") as f: #write
|
37 |
+
f.write(str("time, message, text\n")) # one time only to get column headers for CSV file
|
38 |
+
f.write(str(text.replace("\n"," ")))
|
39 |
+
f.write('\n')
|
40 |
+
return
|
41 |
|
|
|
|
|
|
|
42 |
|
43 |
+
def store_message(name: str, message: str, outputfileName: str):
|
44 |
+
basedir = os.path.dirname(__file__)
|
45 |
+
savePath = outputfileName
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
|
47 |
+
# if file doesnt exist, create it with labels
|
48 |
+
from os.path import exists
|
49 |
+
file_exists = exists(savePath)
|
50 |
+
|
51 |
+
if (file_exists==False):
|
52 |
+
with open(savePath, "w") as f: #write
|
53 |
+
f.write(str("time, message, text\n")) # one time only to get column headers for CSV file
|
54 |
+
if name and message:
|
55 |
+
writer = csv.DictWriter(f, fieldnames=["time", "message", "name"])
|
56 |
+
writer.writerow(
|
57 |
+
{"time": str(datetime.now()), "message": message.strip(), "name": name.strip() }
|
58 |
+
)
|
59 |
+
df = pd.read_csv(savePath)
|
60 |
+
df = df.sort_values(df.columns[0],ascending=False)
|
61 |
+
else:
|
62 |
+
if name and message:
|
63 |
+
with open(savePath, "a") as csvfile:
|
64 |
+
writer = csv.DictWriter(csvfile, fieldnames=[ "time", "message", "name", ])
|
65 |
+
writer.writerow(
|
66 |
+
{"time": str(datetime.now()), "message": message.strip(), "name": name.strip() }
|
67 |
+
)
|
68 |
+
df = pd.read_csv(savePath)
|
69 |
+
df = df.sort_values(df.columns[0],ascending=False)
|
70 |
+
return df
|
71 |
+
|
72 |
mname = "facebook/blenderbot-400M-distill"
|
73 |
model = BlenderbotForConditionalGeneration.from_pretrained(mname)
|
74 |
tokenizer = BlenderbotTokenizer.from_pretrained(mname)
|
75 |
|
76 |
def take_last_tokens(inputs, note_history, history):
|
|
|
77 |
if inputs['input_ids'].shape[1] > 128:
|
78 |
inputs['input_ids'] = torch.tensor([inputs['input_ids'][0][-128:].tolist()])
|
79 |
inputs['attention_mask'] = torch.tensor([inputs['attention_mask'][0][-128:].tolist()])
|
|
|
82 |
return inputs, note_history, history
|
83 |
|
84 |
def add_note_to_history(note, note_history):# good example of non async since we wait around til we know it went okay.
|
|
|
85 |
note_history.append(note)
|
86 |
note_history = '</s> <s>'.join(note_history)
|
87 |
return [note_history]
|
|
|
90 |
description = """Chatbot With persistent memory dataset allowing multiagent system AI to access a shared dataset as memory pool with stored interactions.
|
91 |
Current Best SOTA Chatbot: https://huggingface.co/facebook/blenderbot-400M-distill?text=Hey+my+name+is+ChatBack%21+Are+you+ready+to+rock%3F """
|
92 |
|
93 |
+
def get_base(filename):
|
94 |
+
basedir = os.path.dirname(__file__)
|
95 |
+
print(basedir)
|
96 |
+
#loadPath = basedir + "\\" + filename # works on windows
|
97 |
+
loadPath = basedir + filename
|
98 |
+
print(loadPath)
|
99 |
+
return loadPath
|
100 |
+
|
101 |
def chat(message, history):
|
102 |
history = history or []
|
103 |
if history:
|
104 |
history_useful = ['</s> <s>'.join([str(a[0])+'</s> <s>'+str(a[1]) for a in history])]
|
105 |
else:
|
106 |
history_useful = []
|
107 |
+
|
108 |
history_useful = add_note_to_history(message, history_useful)
|
109 |
inputs = tokenizer(history_useful, return_tensors="pt")
|
110 |
inputs, history_useful, history = take_last_tokens(inputs, history_useful, history)
|
|
|
117 |
df=pd.DataFrame()
|
118 |
|
119 |
if UseMemory:
|
120 |
+
#outputfileName = 'ChatbotMemory.csv'
|
121 |
+
outputfileName = 'ChatbotMemory3.csv' # Test first time file create
|
122 |
+
df = store_message(message, response, outputfileName) # Save to dataset
|
123 |
+
basedir = get_base(outputfileName)
|
124 |
+
|
125 |
+
return history, df, basedir
|
|
|
|
|
126 |
|
127 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
128 |
with gr.Blocks() as demo:
|
129 |
+
gr.Markdown("<h1><center>🍰Gradio chatbot backed by dataframe CSV memory🎨</center></h1>")
|
|
|
130 |
|
131 |
with gr.Row():
|
132 |
t1 = gr.Textbox(lines=1, default="", label="Chat Text:")
|
133 |
+
b1 = gr.Button("Respond and Retrieve Messages")
|
134 |
|
135 |
with gr.Row(): # inputs and buttons
|
136 |
s1 = gr.State([])
|
|
|
|
|
137 |
df1 = gr.Dataframe(wrap=True, max_rows=1000, overflow_row_behaviour= "paginate")
|
138 |
+
with gr.Row(): # inputs and buttons
|
139 |
+
file = gr.File(label="File")
|
140 |
+
s2 = gr.Markdown()
|
141 |
|
142 |
+
b1.click(fn=chat, inputs=[t1, s1], outputs=[s1, df1, file])
|
143 |
+
|
144 |
+
demo.launch(debug=True, show_error=True)
|