update train hparams
This commit is contained in:
parent
4dc0632145
commit
dc4a00dd63
|
@ -37,5 +37,5 @@ pure_bf16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -38,5 +38,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -38,5 +38,5 @@ pure_bf16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -36,5 +36,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -35,5 +35,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -35,5 +35,5 @@ pure_bf16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -37,5 +37,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -37,5 +37,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -38,5 +38,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -38,5 +38,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -36,5 +36,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -34,5 +34,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -33,5 +33,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -34,5 +34,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -34,5 +34,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -35,5 +35,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -34,5 +34,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -34,5 +34,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -35,5 +35,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -34,5 +34,5 @@ fp16: true
|
|||
### eval
|
||||
val_size: 0.1
|
||||
per_device_eval_batch_size: 1
|
||||
evaluation_strategy: steps
|
||||
eval_strategy: steps
|
||||
eval_steps: 500
|
||||
|
|
|
@ -51,4 +51,4 @@ def print_env() -> None:
|
|||
|
||||
info["vLLM version"] = vllm.__version__
|
||||
|
||||
print("\n".join(["- {}: {}".format(key, value) for key, value in info.items()]) + "\n")
|
||||
print("\n" + "\n".join(["- {}: {}".format(key, value) for key, value in info.items()]) + "\n")
|
||||
|
|
|
@ -200,7 +200,7 @@ class Runner:
|
|||
# eval config
|
||||
if get("train.val_size") > 1e-6 and args["stage"] != "ppo":
|
||||
args["val_size"] = get("train.val_size")
|
||||
args["evaluation_strategy"] = "steps"
|
||||
args["eval_strategy"] = "steps"
|
||||
args["eval_steps"] = args["save_steps"]
|
||||
args["per_device_eval_batch_size"] = args["per_device_train_batch_size"]
|
||||
|
||||
|
|
Loading…
Reference in New Issue