Issue in evaluating checkpoints trained by QeRL

[oncleJules]

After training with QeRL, i got such a checkpoint folder

when i tried to use the recommended command
python -m eval.gsm8k --model "~/QeRL/ckpt/qwen2.5-3B-single-gpus-quantized 1e-5_0.28 _gsm8k_b2_g1_r16_True/checkpoint-7473" --load nvfp4
for evaluation, i got the following error

The "Repo id must be in the form 'repo_name'..." issue seems not fixed

Any suggestion to fix this issue? thanks in advance!

[fzhiy]

After training with QeRL, i got such a checkpoint folder

when i tried to use the recommended command `python -m eval.gsm8k --model "~/QeRL/ckpt/qwen2.5-3B-single-gpus-quantized 1e-5_0.28 _gsm8k_b2_g1_r16_True/checkpoint-7473" --load nvfp4` for evaluation, i got the following error The "Repo id must be in the form 'repo_name'..." issue seems not fixed

Any suggestion to fix this issue? thanks in advance!

A relative model path "ckpt/qwen2.5-3B-single-gpus-quantized 1e-5_0.28 _gsm8k_b2_g1_r16_True/checkpoint-7473" would work.

[oncleJules]

After training with QeRL, i got such a checkpoint folder

when i tried to use the recommended command python -m eval.gsm8k --model "~/QeRL/ckpt/qwen2.5-3B-single-gpus-quantized 1e-5_0.28 _gsm8k_b2_g1_r16_True/checkpoint-7473" --load nvfp4 for evaluation, i got the following error

The "Repo id must be in the form 'repo_name'..." issue seems not fixed
Any suggestion to fix this issue? thanks in advance!

A relative model path "ckpt/qwen2.5-3B-single-gpus-quantized 1e-5_0.28 _gsm8k_b2_g1_r16_True/checkpoint-7473" would work.

Thank you for your reply!

however, i've changed to a relative model path, but the error still showed up