Spaces:
Running
on
Zero
Running
on
Zero
Commit
·
8e536a8
1
Parent(s):
a7c1b7f
Disable field model
Browse files
app.py
CHANGED
@@ -20,6 +20,8 @@ This space simply performs inference on the two pretrained models available as
|
|
20 |
part of the ReSym artifacts. It takes a variable name and some decompiled code
|
21 |
as input, and outputs the variable type and other information.
|
22 |
|
|
|
|
|
23 |
## Disclaimer
|
24 |
|
25 |
I'm not a ReSym developer and I may have messed something up. In particular,
|
@@ -39,9 +41,9 @@ tokenizer = AutoTokenizer.from_pretrained("bigcode/starcoderbase-3b")
|
|
39 |
vardecoder_model = AutoModelForCausalLM.from_pretrained(
|
40 |
"ejschwartz/resym-vardecoder", torch_dtype=torch.bfloat16, device_map="auto"
|
41 |
)
|
42 |
-
fielddecoder_model = AutoModelForCausalLM.from_pretrained(
|
43 |
-
|
44 |
-
)
|
45 |
|
46 |
example = r"""__int64 __fastcall sub_410D81(__int64 a1, __int64 a2, __int64 a3)
|
47 |
{
|
@@ -103,24 +105,24 @@ def infer(code):
|
|
103 |
skip_special_tokens=True,
|
104 |
clean_up_tokenization_spaces=True,
|
105 |
)
|
106 |
-
field_output = fielddecoder_model.generate(
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
)[0]
|
116 |
-
field_output = tokenizer.decode(
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
)
|
121 |
|
122 |
var_output = var_name + ":" + var_output
|
123 |
-
field_output = var_name + ":" + field_output
|
124 |
return var_output, varstring
|
125 |
|
126 |
|
|
|
20 |
part of the ReSym artifacts. It takes a variable name and some decompiled code
|
21 |
as input, and outputs the variable type and other information.
|
22 |
|
23 |
+
The examples are randomly selected from `vardecoder_test.jsonl`.
|
24 |
+
|
25 |
## Disclaimer
|
26 |
|
27 |
I'm not a ReSym developer and I may have messed something up. In particular,
|
|
|
41 |
vardecoder_model = AutoModelForCausalLM.from_pretrained(
|
42 |
"ejschwartz/resym-vardecoder", torch_dtype=torch.bfloat16, device_map="auto"
|
43 |
)
|
44 |
+
# fielddecoder_model = AutoModelForCausalLM.from_pretrained(
|
45 |
+
# "ejschwartz/resym-fielddecoder", torch_dtype=torch.bfloat16, device_map="auto"
|
46 |
+
# )
|
47 |
|
48 |
example = r"""__int64 __fastcall sub_410D81(__int64 a1, __int64 a2, __int64 a3)
|
49 |
{
|
|
|
105 |
skip_special_tokens=True,
|
106 |
clean_up_tokenization_spaces=True,
|
107 |
)
|
108 |
+
# field_output = fielddecoder_model.generate(
|
109 |
+
# input_ids=input_ids,
|
110 |
+
# max_new_tokens=1024,
|
111 |
+
# num_beams=4,
|
112 |
+
# num_return_sequences=1,
|
113 |
+
# do_sample=False,
|
114 |
+
# early_stopping=False,
|
115 |
+
# pad_token_id=0,
|
116 |
+
# eos_token_id=0,
|
117 |
+
# )[0]
|
118 |
+
# field_output = tokenizer.decode(
|
119 |
+
# field_output[input_ids.size(1) :],
|
120 |
+
# skip_special_tokens=True,
|
121 |
+
# clean_up_tokenization_spaces=True,
|
122 |
+
# )
|
123 |
|
124 |
var_output = var_name + ":" + var_output
|
125 |
+
#field_output = var_name + ":" + field_output
|
126 |
return var_output, varstring
|
127 |
|
128 |
|