Spaces:
Running
Running
code clean up
Browse files
app.py
CHANGED
@@ -24,11 +24,13 @@ try:
|
|
24 |
gr.Info("Setting up retriever, please wait...")
|
25 |
rag_initial_output = RAG.search("what is Mistral?", k = 1)
|
26 |
gr.Info("Retriever working successfully!")
|
|
|
27 |
except:
|
28 |
gr.Warning("Retriever not working!")
|
29 |
|
30 |
mark_text = '# 🔍 Search Results\n'
|
31 |
header_text = "# ArXivCS RAG \n"
|
|
|
32 |
try:
|
33 |
with open("README.md", "r") as f:
|
34 |
mdfile = f.read()
|
@@ -37,6 +39,7 @@ try:
|
|
37 |
date = match.group().split(': ')[1]
|
38 |
formatted_date = datetime.strptime(date, '%Y-%m-%d').strftime('%d %b %Y')
|
39 |
header_text += f'Index Last Updated: {formatted_date}\n'
|
|
|
40 |
except:
|
41 |
pass
|
42 |
|
@@ -45,6 +48,7 @@ if show_examples:
|
|
45 |
sample_outputs = json.load(f)
|
46 |
output_placeholder = sample_outputs['output_placeholder']
|
47 |
md_text_initial = sample_outputs['search_placeholder']
|
|
|
48 |
else:
|
49 |
output_placeholder = None
|
50 |
md_text_initial = ''
|
@@ -61,6 +65,7 @@ def get_prompt_text(question, context, formatted = True, llm_model_picked = 'mis
|
|
61 |
if formatted:
|
62 |
sys_instruction = f"Context:\n {context} \n Given the following scientific paper abstracts, take a deep breath and lets think step by step to answer the question. Cite the titles of your sources when answering, do not cite links or dates."
|
63 |
message = f"Question: {question}"
|
|
|
64 |
if 'mistralai' in llm_model_picked:
|
65 |
return f"<s>" + f"[INST] {sys_instruction}" + f" {message}[/INST]"
|
66 |
|
@@ -74,12 +79,14 @@ def get_references(question, retriever, k = retrieve_results):
|
|
74 |
return rag_out
|
75 |
|
76 |
def get_rag(message):
|
77 |
-
|
78 |
|
79 |
with gr.Blocks(theme = gr.themes.Soft()) as demo:
|
80 |
header = gr.Markdown(header_text)
|
|
|
81 |
with gr.Group():
|
82 |
msg = gr.Textbox(label = 'Search', placeholder = 'What is Mistral?')
|
|
|
83 |
with gr.Accordion("Advanced Settings", open=False):
|
84 |
with gr.Row(equal_height = True):
|
85 |
llm_model = gr.Dropdown(choices = llm_models_to_choose, value = 'mistralai/Mistral-7B-Instruct-v0.2', label = 'LLM Model')
|
@@ -97,7 +104,6 @@ with gr.Blocks(theme = gr.themes.Soft()) as demo:
|
|
97 |
rag_answer = rag_out[i]
|
98 |
title = rag_answer['document_metadata']['title'].replace('\n','')
|
99 |
|
100 |
-
#score = round(rag_answer['score'], 2)
|
101 |
date = rag_answer['document_metadata']['_time']
|
102 |
paper_title = f'''### {date} | [{title}](https://arxiv.org/abs/{rag_answer['document_id']}) | [⬇️](https://arxiv.org/pdf/{rag_answer['document_id']})\n'''
|
103 |
paper_abs = rag_answer['content']
|
@@ -111,6 +117,7 @@ with gr.Blocks(theme = gr.themes.Soft()) as demo:
|
|
111 |
def ask_llm(prompt, llm_model_picked = 'mistralai/Mistral-7B-Instruct-v0.2', stream_outputs = False):
|
112 |
model_disabled_text = "LLM Model is disabled"
|
113 |
output = ""
|
|
|
114 |
if llm_model_picked == 'None':
|
115 |
if stream_outputs:
|
116 |
for out in model_disabled_text:
|
@@ -123,10 +130,10 @@ with gr.Blocks(theme = gr.themes.Soft()) as demo:
|
|
123 |
client = InferenceClient(llm_model_picked)
|
124 |
try:
|
125 |
stream = client.text_generation(prompt, **generate_kwargs, stream=stream_outputs, details=False, return_full_text=False)
|
|
|
126 |
except:
|
127 |
gr.Warning("LLM Inference rate limit reached, try again later!")
|
128 |
return ""
|
129 |
-
#output = output.lstrip(' \n') if output.lstrip().startswith('\n') else output
|
130 |
|
131 |
if stream_outputs:
|
132 |
for response in stream:
|
@@ -139,4 +146,4 @@ with gr.Blocks(theme = gr.themes.Soft()) as demo:
|
|
139 |
|
140 |
msg.submit(update_with_rag_md, [msg, llm_results, llm_model], [gr_md, input]).success(ask_llm, [input, llm_model, stream_results], output_text)
|
141 |
|
142 |
-
demo.queue(
|
|
|
24 |
gr.Info("Setting up retriever, please wait...")
|
25 |
rag_initial_output = RAG.search("what is Mistral?", k = 1)
|
26 |
gr.Info("Retriever working successfully!")
|
27 |
+
|
28 |
except:
|
29 |
gr.Warning("Retriever not working!")
|
30 |
|
31 |
mark_text = '# 🔍 Search Results\n'
|
32 |
header_text = "# ArXivCS RAG \n"
|
33 |
+
|
34 |
try:
|
35 |
with open("README.md", "r") as f:
|
36 |
mdfile = f.read()
|
|
|
39 |
date = match.group().split(': ')[1]
|
40 |
formatted_date = datetime.strptime(date, '%Y-%m-%d').strftime('%d %b %Y')
|
41 |
header_text += f'Index Last Updated: {formatted_date}\n'
|
42 |
+
|
43 |
except:
|
44 |
pass
|
45 |
|
|
|
48 |
sample_outputs = json.load(f)
|
49 |
output_placeholder = sample_outputs['output_placeholder']
|
50 |
md_text_initial = sample_outputs['search_placeholder']
|
51 |
+
|
52 |
else:
|
53 |
output_placeholder = None
|
54 |
md_text_initial = ''
|
|
|
65 |
if formatted:
|
66 |
sys_instruction = f"Context:\n {context} \n Given the following scientific paper abstracts, take a deep breath and lets think step by step to answer the question. Cite the titles of your sources when answering, do not cite links or dates."
|
67 |
message = f"Question: {question}"
|
68 |
+
|
69 |
if 'mistralai' in llm_model_picked:
|
70 |
return f"<s>" + f"[INST] {sys_instruction}" + f" {message}[/INST]"
|
71 |
|
|
|
79 |
return rag_out
|
80 |
|
81 |
def get_rag(message):
|
82 |
+
return get_references(message, RAG)
|
83 |
|
84 |
with gr.Blocks(theme = gr.themes.Soft()) as demo:
|
85 |
header = gr.Markdown(header_text)
|
86 |
+
|
87 |
with gr.Group():
|
88 |
msg = gr.Textbox(label = 'Search', placeholder = 'What is Mistral?')
|
89 |
+
|
90 |
with gr.Accordion("Advanced Settings", open=False):
|
91 |
with gr.Row(equal_height = True):
|
92 |
llm_model = gr.Dropdown(choices = llm_models_to_choose, value = 'mistralai/Mistral-7B-Instruct-v0.2', label = 'LLM Model')
|
|
|
104 |
rag_answer = rag_out[i]
|
105 |
title = rag_answer['document_metadata']['title'].replace('\n','')
|
106 |
|
|
|
107 |
date = rag_answer['document_metadata']['_time']
|
108 |
paper_title = f'''### {date} | [{title}](https://arxiv.org/abs/{rag_answer['document_id']}) | [⬇️](https://arxiv.org/pdf/{rag_answer['document_id']})\n'''
|
109 |
paper_abs = rag_answer['content']
|
|
|
117 |
def ask_llm(prompt, llm_model_picked = 'mistralai/Mistral-7B-Instruct-v0.2', stream_outputs = False):
|
118 |
model_disabled_text = "LLM Model is disabled"
|
119 |
output = ""
|
120 |
+
|
121 |
if llm_model_picked == 'None':
|
122 |
if stream_outputs:
|
123 |
for out in model_disabled_text:
|
|
|
130 |
client = InferenceClient(llm_model_picked)
|
131 |
try:
|
132 |
stream = client.text_generation(prompt, **generate_kwargs, stream=stream_outputs, details=False, return_full_text=False)
|
133 |
+
|
134 |
except:
|
135 |
gr.Warning("LLM Inference rate limit reached, try again later!")
|
136 |
return ""
|
|
|
137 |
|
138 |
if stream_outputs:
|
139 |
for response in stream:
|
|
|
146 |
|
147 |
msg.submit(update_with_rag_md, [msg, llm_results, llm_model], [gr_md, input]).success(ask_llm, [input, llm_model, stream_results], output_text)
|
148 |
|
149 |
+
demo.queue().launch()
|