first inter-domain generalisation experiments (#412)

* tweak len reward

* first inter-generalisation experiment config

* update inter algorithmic config

* default to empty config

* fix typo

* change config to match experiment script

* long prompt fixes

* algorithmic training config tweaks

* imports

* update algorithmic training cfgs

* first logic composite config

* fix dset name

* tweaks

* fix syllogisms dataset

* rm temp print

* initial algebra config

* algebra cfg tweaks

* add gc

* add initial games cfg

* rename games cfg

* fix dset name

* fix sokoban metadata

* remove boxnet

* games cfg tweak
This commit is contained in:
Oliver Stanley 2025-04-14 21:06:40 +01:00 committed by GitHub
parent ff5407f766
commit 224532f12a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
10 changed files with 934 additions and 23 deletions

View file

@ -100,17 +100,26 @@ def make_dataset(
tokenizer,
data_source: Experiment | ProceduralDataset,
developer_prompt: str,
max_prompt_length: int = 2048,
) -> ReasoningGymDataset:
"""
Create ReasoningGymDataset object using either a ProceduralDataset or Experiment as the underlying data source.
"""
kwargs = {
"tokenizer": tokenizer,
# "dataset_name": dataset_name,
"developer_prompt": developer_prompt,
}
if isinstance(data_source, Experiment):
kwargs["experiment"] = data_source
return ReasoningGymDataset(
tokenizer=tokenizer,
experiment=data_source,
developer_prompt=developer_prompt,
developer_role="system",
max_prompt_length=max_prompt_length,
truncation="error",
)
else:
kwargs["procedural_dataset"] = data_source
return ReasoningGymDataset(**kwargs)
return ReasoningGymDataset(
tokenizer=tokenizer,
procedural_dataset=data_source,
developer_prompt=developer_prompt,
developer_role="system",
max_prompt_length=max_prompt_length,
truncation="error",
)