Update README.md
Browse files
README.md
CHANGED
|
@@ -1,16 +1,31 @@
|
|
| 1 |
DATASET = "task-focus + sample from remain datasets"
|
|
|
|
| 2 |
DATASET_FORMAT = 'input-output'
|
|
|
|
| 3 |
PER_DEVICE_TRAIN_BATCH_SIZE = 2
|
|
|
|
| 4 |
GRADIENT_ACCUMULATION_STEPS = 4
|
|
|
|
| 5 |
LEARNING_RATE = 0.0003
|
|
|
|
| 6 |
LR_SCHEDULER_TYPE = 'constant'
|
|
|
|
| 7 |
WARMUP_RATIO = 0.03
|
|
|
|
| 8 |
LORA_R = 128
|
|
|
|
| 9 |
LORA_ALPHA = 32
|
|
|
|
| 10 |
LORA_DROPOUT = 0.1
|
|
|
|
| 11 |
TRAIN_ON_SOURCE = False
|
|
|
|
| 12 |
SOURCE_MAX_LENGTH = 1024
|
|
|
|
| 13 |
TARGET_MAX_LENGTH = 1024
|
|
|
|
| 14 |
LOGGING_STEPS = 20
|
|
|
|
| 15 |
SAVE_STEPS = 100
|
|
|
|
| 16 |
SAVE_TOTAL_LIMIT = 4
|
|
|
|
| 1 |
DATASET = "task-focus + sample from remain datasets"
|
| 2 |
+
|
| 3 |
DATASET_FORMAT = 'input-output'
|
| 4 |
+
|
| 5 |
PER_DEVICE_TRAIN_BATCH_SIZE = 2
|
| 6 |
+
|
| 7 |
GRADIENT_ACCUMULATION_STEPS = 4
|
| 8 |
+
|
| 9 |
LEARNING_RATE = 0.0003
|
| 10 |
+
|
| 11 |
LR_SCHEDULER_TYPE = 'constant'
|
| 12 |
+
|
| 13 |
WARMUP_RATIO = 0.03
|
| 14 |
+
|
| 15 |
LORA_R = 128
|
| 16 |
+
|
| 17 |
LORA_ALPHA = 32
|
| 18 |
+
|
| 19 |
LORA_DROPOUT = 0.1
|
| 20 |
+
|
| 21 |
TRAIN_ON_SOURCE = False
|
| 22 |
+
|
| 23 |
SOURCE_MAX_LENGTH = 1024
|
| 24 |
+
|
| 25 |
TARGET_MAX_LENGTH = 1024
|
| 26 |
+
|
| 27 |
LOGGING_STEPS = 20
|
| 28 |
+
|
| 29 |
SAVE_STEPS = 100
|
| 30 |
+
|
| 31 |
SAVE_TOTAL_LIMIT = 4
|