georeactor commited on
Commit
7fbdf32
·
1 Parent(s): db91e40

CodeParrot main model

Browse files
Files changed (1) hide show
  1. app.py +11 -6
app.py CHANGED
@@ -18,10 +18,10 @@ def rename_customer(id, newName):\n\t# PROMPT\n\tcur.execute("UPDATE customer SE
18
 
19
  modelPath = {
20
  # "GPT2-Medium": "gpt2-medium",
21
- "CodeParrot-small": "codeparrot/codeparrot-small",
22
  # "CodeGen-350-Mono": "Salesforce/codegen-350M-mono",
23
  # "GPT-Neo-1.3B": "EleutherAI/gpt-neo-1.3B",
24
- # "CodeParrot": "codeparrot/codeparrot",
25
  # "CodeGen-2B-Mono": "Salesforce/codegen-2B-mono",
26
  }
27
 
@@ -29,6 +29,9 @@ preloadModels = {}
29
  for m in list(modelPath.keys()):
30
  preloadModels[m] = ecco.from_pretrained(modelPath[m])
31
 
 
 
 
32
  def generation(tokenizer, model, content):
33
  decoder = 'Standard'
34
  num_beams = 2 if decoder == 'Beam' else None
@@ -63,6 +66,7 @@ def generation(tokenizer, model, content):
63
  return [
64
  "".join(full_output.tokens),
65
  str(prob.item() * 100),
 
66
  ]
67
 
68
  def clean_comment(txt):
@@ -78,8 +82,8 @@ def code_from_prompts(
78
  # tokenizer = AutoTokenizer.from_pretrained(modelPath[model])
79
  # model = ecco.from_pretrained(modelPath[model])
80
  # model = preloadModels[model]
81
- tokenizer = AutoTokenizer.from_pretrained(modelPath["CodeParrot-small"])
82
- model = preloadModels["CodeParrot-small"]
83
 
84
  code = ""
85
  headerComment = headerComment.strip()
@@ -108,7 +112,7 @@ def get_customer(id):\n\tcur.execute('SELECT * FROM customers WHERE id = %s', st
108
  prob = float(results[1])
109
  requests.post("https://code-adv.herokuapp.com/dbpost", json={
110
  "password": os.environ.get('SERVER_PASS', 'help'),
111
- "model": "codeparrot/codeparrot-small",
112
  "headerComment": headerComment,
113
  "bodyComment": fnComment,
114
  "prefunction": pre_content,
@@ -134,7 +138,8 @@ iface = gr.Interface(
134
  outputs=[
135
  gr.components.Textbox(label="Most probable code"),
136
  gr.components.Textbox(label="Probability of concat"),
 
137
  ],
138
- description="Prompt the code model to write a SQL query with string concatenation - Evaluation on CodeParrot-small - leaderboard coming at https://code-adv.herokuapp.com/dbcompose",
139
  )
140
  iface.launch()
 
18
 
19
  modelPath = {
20
  # "GPT2-Medium": "gpt2-medium",
21
+ # "CodeParrot-small": "codeparrot/codeparrot-small",
22
  # "CodeGen-350-Mono": "Salesforce/codegen-350M-mono",
23
  # "GPT-Neo-1.3B": "EleutherAI/gpt-neo-1.3B",
24
+ "CodeParrot": "codeparrot/codeparrot",
25
  # "CodeGen-2B-Mono": "Salesforce/codegen-2B-mono",
26
  }
27
 
 
29
  for m in list(modelPath.keys()):
30
  preloadModels[m] = ecco.from_pretrained(modelPath[m])
31
 
32
+ topComments = []
33
+ rankings = requests.get("https://code-adv.herokuapp.com/db").json()['results']
34
+
35
  def generation(tokenizer, model, content):
36
  decoder = 'Standard'
37
  num_beams = 2 if decoder == 'Beam' else None
 
66
  return [
67
  "".join(full_output.tokens),
68
  str(prob.item() * 100),
69
+ rankings
70
  ]
71
 
72
  def clean_comment(txt):
 
82
  # tokenizer = AutoTokenizer.from_pretrained(modelPath[model])
83
  # model = ecco.from_pretrained(modelPath[model])
84
  # model = preloadModels[model]
85
+ tokenizer = AutoTokenizer.from_pretrained(modelPath["CodeParrot"])
86
+ model = preloadModels["CodeParrot"]
87
 
88
  code = ""
89
  headerComment = headerComment.strip()
 
112
  prob = float(results[1])
113
  requests.post("https://code-adv.herokuapp.com/dbpost", json={
114
  "password": os.environ.get('SERVER_PASS', 'help'),
115
+ "model": "codeparrot/codeparrot",
116
  "headerComment": headerComment,
117
  "bodyComment": fnComment,
118
  "prefunction": pre_content,
 
138
  outputs=[
139
  gr.components.Textbox(label="Most probable code"),
140
  gr.components.Textbox(label="Probability of concat"),
141
+ gr.components.Json(value=rankings)
142
  ],
143
+ description="Prompt the code model to write a SQL query with string concatenation - Evaluation on CodeParrot - leaderboard coming at https://code-adv.herokuapp.com/dbcompose",
144
  )
145
  iface.launch()