ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'main' into nightly

Browse source
This commit is contained in:
Daniel Han 2025-06-22 02:37:11 -07:00
commit 2cb0ebc47e
1 changed files with 7 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

8
unsloth/models/rl.py
View file

@ -236,7 +236,9 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
if "args" in call_args and "model" in call_args:
mixed_precision = \
"use_bf16 = getattr(args, 'bf16', False)\n"\
"if type(use_bf16) is not bool: use_bf16 = False\n"\
"use_fp16 = getattr(args, 'fp16', False)\n"\
"if type(use_fp16) is not bool: use_fp16 = False\n"\
"force_float32 = False\n"\
"if os.environ.get('UNSLOTH_FORCE_FLOAT32', '0') == '1':\n"\
" print('Unsloth: Switching to float32 training since model cannot work with float16')\n"\
@ -293,7 +295,9 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
" if eval_bsz == 8 and args.per_device_train_batch_size < eval_bsz: args.per_device_eval_batch_size = args.per_device_train_batch_size\n"\
" if getattr(args, 'eval_accumulation_steps', None) is None and ga_steps is not None: args.eval_accumulation_steps = ga_steps\n"\
"fp16_full_eval = getattr(args, 'fp16_full_eval', False)\n"\
"if type(fp16_full_eval) is not bool: fp16_full_eval = False\n"\
"bf16_full_eval = getattr(args, 'bf16_full_eval', False)\n"\
"if type(bf16_full_eval) is not bool: bf16_full_eval = False\n"\
"if args.fp16 and bf16_full_eval: args.bf16_full_eval = False; args.fp16_full_eval = True\n"\
"if args.bf16 and fp16_full_eval: args.bf16_full_eval = True; args.fp16_full_eval = False\n"\
"if force_float32:\n"\
@ -472,6 +476,8 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
"top_k" : None,
"vllm_mode" : "colocate",
"generation_kwargs" : {},
"bf16" : False,
"fp16" : False,
}
for k, v in replacements.items():
x = f"{k}( = [^,\n]{{1,}})?,\n"
@ -531,7 +537,7 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
" mask_truncated_completions = True\n"\
" epsilon_high = 0.28\n"\
" beta = 0.0\n"\
" loss_type = 'bnpo'"
" loss_type = 'bnpo'\n"\
"\n"
extra_args += check_dr_grpo
pass