pretrain model
Browse files
scripts/pretrain-model.yaml
CHANGED
@@ -52,7 +52,7 @@ data:
|
|
52 |
# Training-related arguments. See ``litgpt.args.TrainArgs`` for details
|
53 |
train:
|
54 |
# Number of optimizer steps between saving checkpoints (type: Optional[int], default: 1000)
|
55 |
-
save_interval:
|
56 |
|
57 |
# Number of iterations between logging calls (type: int, default: 1)
|
58 |
log_interval: 1
|
@@ -77,7 +77,7 @@ train:
|
|
77 |
max_steps:
|
78 |
|
79 |
# Limits the length of samples. Off by default (type: Optional[int], default: null)
|
80 |
-
max_seq_length:
|
81 |
|
82 |
# Whether to tie the embedding weights with the language modeling head weights. (type: Optional[bool], default: False)
|
83 |
tie_embeddings:
|
@@ -91,7 +91,7 @@ train:
|
|
91 |
# Evaluation-related arguments. See ``litgpt.args.EvalArgs`` for details
|
92 |
eval:
|
93 |
# Number of optimizer steps between evaluation calls (type: int, default: 1000)
|
94 |
-
interval:
|
95 |
|
96 |
# Number of tokens to generate (type: Optional[int], default: null)
|
97 |
max_new_tokens:
|
|
|
52 |
# Training-related arguments. See ``litgpt.args.TrainArgs`` for details
|
53 |
train:
|
54 |
# Number of optimizer steps between saving checkpoints (type: Optional[int], default: 1000)
|
55 |
+
save_interval: 500
|
56 |
|
57 |
# Number of iterations between logging calls (type: int, default: 1)
|
58 |
log_interval: 1
|
|
|
77 |
max_steps:
|
78 |
|
79 |
# Limits the length of samples. Off by default (type: Optional[int], default: null)
|
80 |
+
max_seq_length:
|
81 |
|
82 |
# Whether to tie the embedding weights with the language modeling head weights. (type: Optional[bool], default: False)
|
83 |
tie_embeddings:
|
|
|
91 |
# Evaluation-related arguments. See ``litgpt.args.EvalArgs`` for details
|
92 |
eval:
|
93 |
# Number of optimizer steps between evaluation calls (type: int, default: 1000)
|
94 |
+
interval: 100
|
95 |
|
96 |
# Number of tokens to generate (type: Optional[int], default: null)
|
97 |
max_new_tokens:
|