Update README.md
Browse files
README.md
CHANGED
@@ -64,32 +64,32 @@ metrics:
|
|
64 |
# Training parameters and hyperparameters
|
65 |
|
66 |
|
67 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
|
69 |
-
|
70 |
|
71 |
-
|
72 |
|
73 |
-
|
74 |
-
|
75 |
-
|
76 |
-
# For training arguments:
|
77 |
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
warmup_steps=50
|
87 |
-
|
88 |
-
weight_decay=0.05
|
89 |
-
|
90 |
-
learning_rate=3e-3
|
91 |
-
|
92 |
-
lr_scheduler_type="linear"
|
93 |
|
94 |
|
95 |
# Training Results
|
|
|
64 |
# Training parameters and hyperparameters
|
65 |
|
66 |
|
67 |
+
The following were used during training:
|
68 |
+
|
69 |
+
# For Lora:
|
70 |
+
|
71 |
+
r=18
|
72 |
+
|
73 |
+
alpha=8
|
74 |
+
|
75 |
+
|
76 |
+
# For training arguments:
|
77 |
+
|
78 |
+
gradient_accumulation_steps=24
|
79 |
|
80 |
+
per_device_train_batch_size=8
|
81 |
|
82 |
+
per_device_eval_batch_size=8
|
83 |
|
84 |
+
max_steps=1000
|
|
|
|
|
|
|
85 |
|
86 |
+
warmup_steps=50
|
87 |
+
|
88 |
+
weight_decay=0.05
|
89 |
+
|
90 |
+
learning_rate=3e-3
|
91 |
+
|
92 |
+
lr_scheduler_type="linear"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
|
94 |
|
95 |
# Training Results
|