Unverified Commit a5f9fb59 authored by wang.yuqi's avatar wang.yuqi Committed by GitHub
Browse files

[Deprecation] Deprecation `--convert reward`, use `--convert embed` instead. (#30463)


Signed-off-by: default avatarwang.yuqi <yuqi.wang@daocloud.io>
parent 4515eb1a
......@@ -316,10 +316,13 @@ We have split the `encode` task into two more specific token-wise tasks: `token_
### Remove softmax from PoolingParams
We are going to remove `softmax` and `activation` from `PoolingParams`. Instead, use `use_activation`, since we allow `classify` and `token_classify` to use any activation function.
We are going to remove `softmax` and `activation` from `PoolingParams` in v0.15. Instead, use `use_activation`, since we allow `classify` and `token_classify` to use any activation function.
### as_reward_model
!!! warning
We are going to remove `--convert reward` in v0.15, use `--convert embed` instead.
Pooling models now default support all pooling, you can use it without any settings.
- Extracting hidden states prefers using `token_embed` task.
......
......@@ -788,6 +788,13 @@ class ModelConfig:
runner_type: RunnerType,
convert: ConvertOption,
) -> ConvertType:
if convert == "reward":
logger.warning(
"`--convert reward` is deprecated and will be removed in v0.15. "
"Please use `--convert embed` instead."
)
return "embed"
if convert != "auto":
return convert
......
......@@ -111,13 +111,15 @@ class PoolerConfig:
def get_use_activation(o: object):
if softmax := getattr(o, "softmax", None) is not None:
logger.warning_once(
"softmax will be deprecated, please use use_activation instead."
"softmax will be deprecated and will be removed in v0.15. "
"Please use use_activation instead."
)
return softmax
if activation := getattr(o, "activation", None) is not None:
logger.warning_once(
"activation will be deprecated, please use use_activation instead."
"activation will be deprecated and will be removed in v0.15. "
"Please use use_activation instead."
)
return activation
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment