ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix RM save model

Browse Source 
Former-commit-id: 8104cc2425431eb1cddccf3909855296116f922b
This commit is contained in:
hiyouga
2023-08-01 11:56:17 +08:00
parent 9bba01a033
commit 8e26eb374e
7 changed files with 33 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/llmtuner/tuner/core/parser.py
View File

@@ -85,6 +85,9 @@ def get_train_args(
assert training_args.evaluation_strategy == "no" or (not data_args.streaming), \
"Streaming mode does not support evaluation currently."
assert not (general_args.stage == "ppo" and data_args.streaming), \
"Streaming mode does not suppport PPO training currently."
if model_args.checkpoint_dir is not None:
if finetuning_args.finetuning_type != "lora":
assert len(model_args.checkpoint_dir) == 1, "Only LoRA tuning accepts multiple checkpoints."
@@ -107,8 +110,8 @@ def get_train_args(
training_args.ddp_find_unused_parameters = False
if data_args.max_samples is not None and data_args.streaming:
logger.warning("`max_samples` is incompatible with `streaming`. Disabling streaming mode.")
data_args.streaming = False
logger.warning("`max_samples` is incompatible with `streaming`. Disabling max_samples.")
data_args.max_samples = None
if data_args.dev_ratio > 1e-6 and data_args.streaming:
logger.warning("`dev_ratio` is incompatible with `streaming`. Disabling development set.")