Add more files
Browse files- etlex-utf8.csv +3 -0
- etlex_convert_to_qwen2_finetune.py +128 -0
- lexitron2_etlex_finetune.jsonl +3 -0
- lexitron2_etlex_finetune.qwen2.txt +3 -0
- lexitron2_telex_finetune.jsonl +3 -0
- lexitron2_telex_finetune.qwen2.txt +3 -0
- telex-utf8.csv +3 -0
- telex_convert_to_qwen2_finetune.py +128 -0
etlex-utf8.csv
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1fc84e257a5d44db9b821731f41cc28da8d41054bf96445db7cdf10622cd08ac
|
3 |
+
size 9599159
|
etlex_convert_to_qwen2_finetune.py
ADDED
@@ -0,0 +1,128 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import csv
|
2 |
+
import random
|
3 |
+
import json
|
4 |
+
|
5 |
+
def create_training_data():
|
6 |
+
training_data = []
|
7 |
+
|
8 |
+
# Read CSV file
|
9 |
+
with open('etlex-utf8.csv', 'r', encoding='utf-8') as f:
|
10 |
+
reader = csv.reader(f)
|
11 |
+
word_data = list(reader)
|
12 |
+
|
13 |
+
for row in word_data:
|
14 |
+
if len(row) < 7: # Skip incomplete rows
|
15 |
+
continue
|
16 |
+
|
17 |
+
english_word = row[1]
|
18 |
+
thai_word = row[3]
|
19 |
+
category = row[4]
|
20 |
+
thai_syn = row[5]
|
21 |
+
eng_syn = row[6]
|
22 |
+
eng_ant = row[7]
|
23 |
+
|
24 |
+
if not english_word: # Skip entries without English word
|
25 |
+
continue
|
26 |
+
|
27 |
+
# Create different types of prompts
|
28 |
+
prompt_types = [
|
29 |
+
f"What is the Thai translation of '{english_word}'?",
|
30 |
+
f"How do you say '{english_word}' in Thai?",
|
31 |
+
f"Can you translate '{english_word}' to Thai?",
|
32 |
+
f"What does '{english_word}' mean in Thai?",
|
33 |
+
f"Give me the Thai equivalent of '{english_word}'",
|
34 |
+
f"Please translate '{english_word}' into Thai",
|
35 |
+
f"What's the Thai word for '{english_word}'?",
|
36 |
+
f"I need the Thai translation for '{english_word}'",
|
37 |
+
f"How would you translate '{english_word}' to Thai?",
|
38 |
+
f"Could you tell me what '{english_word}' is in Thai?",
|
39 |
+
f"What is '{english_word}' in Thai language?",
|
40 |
+
f"Translate '{english_word}' from English to Thai",
|
41 |
+
f"Can you explain the meaning of '{english_word}' in Thai?",
|
42 |
+
f"I want to know how to say '{english_word}' in Thai",
|
43 |
+
f"Please provide the Thai translation and usage of '{english_word}'",
|
44 |
+
f"What's the Thai meaning of '{english_word}'?",
|
45 |
+
f"How is '{english_word}' expressed in Thai?",
|
46 |
+
f"Give me the Thai definition of '{english_word}'",
|
47 |
+
f"I'd like to know the Thai equivalent of '{english_word}'",
|
48 |
+
f"What's the Thai word and meaning for '{english_word}'?",
|
49 |
+
f"คำว่า '{english_word}' แปลว่าอะไร",
|
50 |
+
f"'{english_word}' แปลเป็นภาษาไทยว่าอย่างไร",
|
51 |
+
f"ช่วยแปลคำว่า '{english_word}' เป็นภาษาไทยหน่อย",
|
52 |
+
f"'{english_word}' ภาษาไทยคือคำว่าอะไร",
|
53 |
+
f"อยากทราบความหมายภาษาไทยของคำว่า '{english_word}'",
|
54 |
+
f"ขอคำแปลภาษาไทยของ '{english_word}'",
|
55 |
+
f"'{english_word}' มีความหมายในภาษาไทยว่าอย่างไร",
|
56 |
+
f"ช่วยบอกความหมายภาษาไทยของ '{english_word}' หน่อย",
|
57 |
+
f"'{english_word}' ในภาษาไทยแปลว่าอะไร",
|
58 |
+
f"ขอความหมายภาษาไทยของคำว่า '{english_word}'"
|
59 |
+
]
|
60 |
+
|
61 |
+
# Randomly select prompt type
|
62 |
+
prompt = random.choice(prompt_types)
|
63 |
+
|
64 |
+
# Convert category abbreviation to full name
|
65 |
+
category_mapping = {
|
66 |
+
'N': 'Noun',
|
67 |
+
'V': 'Verb',
|
68 |
+
'VT': 'Transitive Verb',
|
69 |
+
'VI': 'Intransitive Verb',
|
70 |
+
'ADJ': 'Adjective',
|
71 |
+
'ADV': 'Adverb',
|
72 |
+
'PREP': 'Preposition',
|
73 |
+
'CONJ': 'Conjunction',
|
74 |
+
'CLAS': 'Classifier',
|
75 |
+
'PRON': 'Pronoun',
|
76 |
+
'DET': 'Determiner',
|
77 |
+
'ABBR': 'Abbreviation'
|
78 |
+
}
|
79 |
+
|
80 |
+
# Map the abbreviated category to full name, default to original if not found
|
81 |
+
category = category_mapping.get(category, category)
|
82 |
+
|
83 |
+
# Create response based on available information
|
84 |
+
response = f"คำว่า '{english_word}' เป็นคำประเภท {category}"
|
85 |
+
|
86 |
+
if thai_word:
|
87 |
+
response += f" แปลเป็นภาษาไทยว่า '{thai_word}'"
|
88 |
+
|
89 |
+
if thai_syn:
|
90 |
+
response += f"\nคำที่มีความหมายเหมือนกันในภาษาไทย: {thai_syn}"
|
91 |
+
|
92 |
+
if eng_ant:
|
93 |
+
response += f"\nคำที่มีความหมายตรงข้ามในภาษาอังกฤษ: {eng_ant}"
|
94 |
+
|
95 |
+
if eng_syn:
|
96 |
+
response += f"\nคำที่มีความหมายเหมือนกันในภาษาอังกฤษ: {eng_syn}"
|
97 |
+
|
98 |
+
# Create data in both formats
|
99 |
+
# Format 1: Qwen2 conversation format
|
100 |
+
conversation = (
|
101 |
+
"<|im_start|>system\n"
|
102 |
+
"คุณคือผู้ช่วยตอบคำถามที่ฉลาดและซื่อสัตย์.<|im_end|>\n"
|
103 |
+
f"<|im_start|>user\n{prompt}<|im_end|>\n"
|
104 |
+
f"<|im_start|>assistant\n{response}<|im_end|>"
|
105 |
+
)
|
106 |
+
|
107 |
+
# Format 2: JSONL format
|
108 |
+
json_data = {
|
109 |
+
"instruction": prompt,
|
110 |
+
"output": response
|
111 |
+
}
|
112 |
+
|
113 |
+
training_data.append((conversation, json_data))
|
114 |
+
|
115 |
+
return training_data
|
116 |
+
|
117 |
+
# Generate training data
|
118 |
+
training_examples = create_training_data()
|
119 |
+
|
120 |
+
# Write to output files
|
121 |
+
with open('lexitron2_etlex_finetune.qwen2.txt', 'w', encoding='utf-8') as f:
|
122 |
+
for example, _ in training_examples:
|
123 |
+
example = example.replace('\n', '\\n')
|
124 |
+
f.write(example + '\n')
|
125 |
+
|
126 |
+
with open('lexitron2_etlex_finetune.jsonl', 'w', encoding='utf-8') as f:
|
127 |
+
for _, json_data in training_examples:
|
128 |
+
f.write(json.dumps(json_data, ensure_ascii=False) + '\n')
|
lexitron2_etlex_finetune.jsonl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:44d1533767a4a84720ad008e1f4cc6c84a44b6c0d7fe6df683c111004918b081
|
3 |
+
size 33944551
|
lexitron2_etlex_finetune.qwen2.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0822f3d39370a7641633ba31720e174a83bac6aa809f8e9806bba04e6c54d0c6
|
3 |
+
size 49176006
|
lexitron2_telex_finetune.jsonl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b9783044ea940fb10e4985372e59f8b3647fab08f27928dfb17aa66c8cf70bd0
|
3 |
+
size 31243813
|
lexitron2_telex_finetune.qwen2.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3415788521911fe59e56f6ad49e64c44010964b0a3e1da8914df4d62a8c19a76
|
3 |
+
size 38678047
|
telex-utf8.csv
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:54e4ae3f5befc2fb0c18ea84f4b6152def64391818fea0dc90df829964b2f38a
|
3 |
+
size 14559356
|
telex_convert_to_qwen2_finetune.py
ADDED
@@ -0,0 +1,128 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import csv
|
2 |
+
import random
|
3 |
+
import json
|
4 |
+
|
5 |
+
def create_training_data():
|
6 |
+
training_data = []
|
7 |
+
|
8 |
+
# Read CSV file
|
9 |
+
with open('telex-utf8.csv', 'r', encoding='utf-8') as f:
|
10 |
+
reader = csv.reader(f)
|
11 |
+
word_data = list(reader)
|
12 |
+
|
13 |
+
for row in word_data:
|
14 |
+
if len(row) < 11: # Skip incomplete rows
|
15 |
+
continue
|
16 |
+
|
17 |
+
thai_word = row[1]
|
18 |
+
english_syn = row[3]
|
19 |
+
category = row[4]
|
20 |
+
thai_syn = row[5]
|
21 |
+
example = row[6]
|
22 |
+
antonym = row[7]
|
23 |
+
definition = row[8]
|
24 |
+
related_eng = row[9]
|
25 |
+
unit_label = row[10]
|
26 |
+
|
27 |
+
if not thai_word: # Skip entries without Thai word
|
28 |
+
continue
|
29 |
+
|
30 |
+
# Create different types of prompts
|
31 |
+
prompt_types = [
|
32 |
+
f"คำว่า '{thai_word}' แปลว่าอะไร",
|
33 |
+
f"ช่วยอธิบายความหมายของคำว่า '{thai_word}'",
|
34 |
+
f"'{thai_word}' มีความหมายว่าอย่างไร",
|
35 |
+
f"ขอตัวอย่างประโยคของคำว่า '{thai_word}'",
|
36 |
+
f"ยกตัวอย่างการใช้คำว่า '{thai_word}'",
|
37 |
+
f"ช่วยอธิบายคำว่า '{thai_word}' หน่อย",
|
38 |
+
f"'{thai_word}' หมายถึงอะไร",
|
39 |
+
f"อยากทราบความหมายของคำว่า '{thai_word}'",
|
40 |
+
f"คำว่า '{thai_word}' ใช้ในบริบทไหนได้บ้าง",
|
41 |
+
f"ขอความหมายของคำว่า '{thai_word}'",
|
42 |
+
f"'{thai_word}' คืออะไร",
|
43 |
+
f"ช่วยยกตัวอย่างการใช้คำว่า '{thai_word}' ในประโยค",
|
44 |
+
f"คำว่า '{thai_word}' สามารถใช้ในประโยคอย่างไรได้บ้าง",
|
45 |
+
f"อยากรู้ว่าคำว่า '{thai_word}' ใช้ในประโยคยังไง",
|
46 |
+
f"ขอทราบความหมายและตัวอย่างการใช้คำว่า '{thai_word}'",
|
47 |
+
f"ช่วยอธิบายความหมายและยกตัวอย่างการใช้คำว่า '{thai_word}'",
|
48 |
+
f"'{thai_word}' มีวิธีใช้อย่างไร",
|
49 |
+
f"คำว่า '{thai_word}' มีความหมายและใช้อย่างไร",
|
50 |
+
f"อยากทราบรายละเอียดเกี่ยวกับคำว่า '{thai_word}'",
|
51 |
+
f"ขอคำอธิบายและตัวอย่างการใช้คำว่า '{thai_word}'"
|
52 |
+
]
|
53 |
+
|
54 |
+
# Randomly select prompt type
|
55 |
+
prompt = random.choice(prompt_types)
|
56 |
+
# Convert category abbreviation to full name
|
57 |
+
category_mapping = {
|
58 |
+
'N': 'Noun (คำนาม)',
|
59 |
+
'V': 'Verb (คำกริยา)',
|
60 |
+
'VT': 'Transitive Verb (คำกริยาที่ต้องการกรรม)',
|
61 |
+
'VI': 'Intransitive Verb (คำกริยาไม่ต้องการกรรม)',
|
62 |
+
'ADJ': 'Adjective (คำคุณศัพท์)',
|
63 |
+
'ADV': 'Adverb (คำวิเศษณ์)',
|
64 |
+
'PREP': 'Preposition (คำบุพบท)',
|
65 |
+
'CONJ': 'Conjunction (คำสันธาน)',
|
66 |
+
'CLAS': 'Classifier (คำลักษณนาม)',
|
67 |
+
'PRON': 'Pronoun (คำสรรพนาม)',
|
68 |
+
'DET': 'Determiner (คำกำหนด)',
|
69 |
+
'ABBR': 'Abbreviation (คำย่อ)'
|
70 |
+
}
|
71 |
+
|
72 |
+
# Map the abbreviated category to full name, default to original if not found
|
73 |
+
category = category_mapping.get(category, category)
|
74 |
+
# Create response based on available information
|
75 |
+
response = f"คำว่า '{thai_word}' เป็นคำประเภท {category}"
|
76 |
+
|
77 |
+
if english_syn:
|
78 |
+
response += f" แปลเป็นภาษาอังกฤษว่า {english_syn}"
|
79 |
+
|
80 |
+
if definition:
|
81 |
+
response += f"\nความหมาย: {definition}"
|
82 |
+
|
83 |
+
if thai_syn:
|
84 |
+
response += f"\nคำที่มีความหมายเหมือนกัน: {thai_syn}"
|
85 |
+
|
86 |
+
if antonym:
|
87 |
+
response += f"\nคำที่มีความหมายตรงข้าม: {antonym}"
|
88 |
+
|
89 |
+
if example:
|
90 |
+
response += f"\nตัวอย่างประโยค: {example}"
|
91 |
+
|
92 |
+
if unit_label:
|
93 |
+
response += f"\nลักษณนาม: {unit_label}"
|
94 |
+
|
95 |
+
if related_eng:
|
96 |
+
response += f"\nคำภาษาอังกฤษที่เกี่ยวข้อง: {related_eng}"
|
97 |
+
|
98 |
+
# Create data in both formats
|
99 |
+
# Format 1: Qwen2 conversation format
|
100 |
+
conversation = (
|
101 |
+
"<|im_start|>system\n"
|
102 |
+
"คุณคือผู้ช่วยตอบคำถามที่ฉลาดและซื่อสัตย์<|im_end|>\n"
|
103 |
+
f"<|im_start|>user\n{prompt}<|im_end|>\n"
|
104 |
+
f"<|im_start|>assistant\n{response}<|im_end|>"
|
105 |
+
)
|
106 |
+
|
107 |
+
# Format 2: JSONL format
|
108 |
+
json_data = {
|
109 |
+
"instruction": prompt,
|
110 |
+
"output": response
|
111 |
+
}
|
112 |
+
|
113 |
+
training_data.append((conversation, json_data))
|
114 |
+
|
115 |
+
return training_data
|
116 |
+
|
117 |
+
# Generate training data
|
118 |
+
training_examples = create_training_data()
|
119 |
+
|
120 |
+
# Write to output files
|
121 |
+
with open('lexitron2_telex_finetune.qwen2.txt', 'w', encoding='utf-8') as f:
|
122 |
+
for example, _ in training_examples:
|
123 |
+
example = example.replace('\n', '\\n')
|
124 |
+
f.write(example + '\n')
|
125 |
+
|
126 |
+
with open('lexitron2_telex_finetune.jsonl', 'w', encoding='utf-8') as f:
|
127 |
+
for _, json_data in training_examples:
|
128 |
+
f.write(json.dumps(json_data, ensure_ascii=False) + '\n')
|