Automatic dcft datacuration framework upload for wia_dcft_webinstruct_original_uniform_50k
Browse files
config/WebInstructAblation/wia_dcft_webinstruct_original_uniform_50k.yaml
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
operators:
|
| 2 |
+
- id: load_stage_3
|
| 3 |
+
config:
|
| 4 |
+
type: hf_source
|
| 5 |
+
dataset: mlfoundations-dev/wia_dcft_stage_3
|
| 6 |
+
split: train
|
| 7 |
+
columns:
|
| 8 |
+
- revised_question
|
| 9 |
+
- revised_answer
|
| 10 |
+
|
| 11 |
+
- id: convert_to_sharegpt
|
| 12 |
+
config:
|
| 13 |
+
type: function
|
| 14 |
+
function: data_strategies.commons.convert_instruction_response_to_sharegpt
|
| 15 |
+
function_config:
|
| 16 |
+
input_instruction_column: revised_question
|
| 17 |
+
input_response_column: revised_answer
|
| 18 |
+
output_sharegpt_column: conversations
|
| 19 |
+
input_ids:
|
| 20 |
+
- load_stage_3
|