Datasets:
Update README.md
Browse filesUpdate example to match current schema.
README.md
CHANGED
@@ -30,7 +30,7 @@ def read_jsonl(file_path):
|
|
30 |
return data
|
31 |
|
32 |
# Read the train and validation files
|
33 |
-
train_data = read_jsonl('training_data/train.jsonl')
|
34 |
valid_data = read_jsonl('training_data/valid.jsonl')
|
35 |
|
36 |
# Convert to pandas DataFrame
|
@@ -67,10 +67,10 @@ def create_prompt(sample):
|
|
67 |
|
68 |
full_prompt = ""
|
69 |
full_prompt += bos_token
|
70 |
-
full_prompt += "[INST] " + system_message + "
|
71 |
full_prompt += question + " [/INST] "
|
72 |
full_prompt += response
|
73 |
-
full_prompt += eos_token
|
74 |
|
75 |
return full_prompt
|
76 |
...
|
|
|
30 |
return data
|
31 |
|
32 |
# Read the train and validation files
|
33 |
+
train_data = read_jsonl('training_data/train.jsonl')
|
34 |
valid_data = read_jsonl('training_data/valid.jsonl')
|
35 |
|
36 |
# Convert to pandas DataFrame
|
|
|
67 |
|
68 |
full_prompt = ""
|
69 |
full_prompt += bos_token
|
70 |
+
full_prompt += "[INST] <<SYS>>" + system_message + "<</SYS>>\n\n"
|
71 |
full_prompt += question + " [/INST] "
|
72 |
full_prompt += response
|
73 |
+
full_prompt += eos_token
|
74 |
|
75 |
return full_prompt
|
76 |
...
|