Upload configuration file
Browse files
config/multiple_samples_majority_consensus_openr1_math_verify.yaml
CHANGED
|
@@ -4,30 +4,31 @@ operators:
|
|
| 4 |
type: hf_source
|
| 5 |
dataset: open-r1/OpenR1-Math-Raw
|
| 6 |
split: train
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
- id: select_rows_without_proof
|
| 8 |
config:
|
| 9 |
type: function
|
| 10 |
sharded: true
|
| 11 |
-
num_shards:
|
| 12 |
function: data_strategies.commons.select_rows_without
|
| 13 |
function_config:
|
| 14 |
conditions:
|
| 15 |
answer: "proof"
|
| 16 |
input_ids:
|
| 17 |
-
-
|
| 18 |
-
- id: sample_dataset
|
| 19 |
-
config:
|
| 20 |
-
type: function
|
| 21 |
-
function: data_strategies.commons.uniform_sample_limited
|
| 22 |
-
function_config:
|
| 23 |
-
num_samples: 100_000
|
| 24 |
-
input_ids:
|
| 25 |
-
- select_rows_without_proof
|
| 26 |
- id: expand_generation
|
| 27 |
config:
|
| 28 |
-
type:
|
|
|
|
| 29 |
sharded: true
|
| 30 |
-
num_shards:
|
| 31 |
function: data_strategies.OpenR1_MultipleSamples.utils.apply_expand_generation
|
| 32 |
input_ids:
|
| 33 |
- sample_dataset
|
|
|
|
| 4 |
type: hf_source
|
| 5 |
dataset: open-r1/OpenR1-Math-Raw
|
| 6 |
split: train
|
| 7 |
+
- id: sample_dataset
|
| 8 |
+
config:
|
| 9 |
+
type: function
|
| 10 |
+
function: data_strategies.commons.uniform_sample_fixed
|
| 11 |
+
function_config:
|
| 12 |
+
num_samples: 100_000
|
| 13 |
+
input_ids:
|
| 14 |
+
- get_openr1_hf
|
| 15 |
- id: select_rows_without_proof
|
| 16 |
config:
|
| 17 |
type: function
|
| 18 |
sharded: true
|
| 19 |
+
num_shards: 64
|
| 20 |
function: data_strategies.commons.select_rows_without
|
| 21 |
function_config:
|
| 22 |
conditions:
|
| 23 |
answer: "proof"
|
| 24 |
input_ids:
|
| 25 |
+
- sample_dataset
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
- id: expand_generation
|
| 27 |
config:
|
| 28 |
+
type: high_memory_function
|
| 29 |
+
memory: 600
|
| 30 |
sharded: true
|
| 31 |
+
num_shards: 64
|
| 32 |
function: data_strategies.OpenR1_MultipleSamples.utils.apply_expand_generation
|
| 33 |
input_ids:
|
| 34 |
- sample_dataset
|