#Trying to train and unsloth studio :One of the subprocesses has abruptly died during map operation.
1 messages ยท Page 1 of 1 (latest)
Hello . we are aware of this problem and working on solving it
i have the same issue i believe
same issue here
imma try a different model and dataset today and see if its still happening on my end.
i was having errors with gemma3-270m and my csv based datasets
@river mica what did your experiment lead to?
@ripe prawn same behaviour
where can i find that?
oh is there an unsloth_studio folder in the directory where you ran the installer?
yes
2026.3.8
im looking in C:\Users\User\unsloth_studio\Lib\site-packages right now
im on windows btw
in the logs
if you look at the log the first user shared
you'll see this
oh wait
wait
qwen3.5 right?
ok
@river mica are you also getting ModuleNotFoundError: No module named 'UnslothSFTTrainer'
the only difference in the traceback is:
AttributeError: Can't get attribute 'unsloth_push_to_hub' on <module 'unsloth.save' from 'C:\\Users\\User\\.unsloth\\studio\\.venv\\Lib\\site-packages\\unsloth\\save.py'> Applying chat template to chatml_conversations (num_proc=1): 0%| | 0/15011 [00:26<?, ? examples/s] {"timestamp": "2026-03-23T08:54:48.175790Z", "level": "error", "event": "Dataset conversion failed: Failed to format ChatML dataset: One of the subprocesses has abruptly died during map operation.To debug the error, disable multiprocessing."} {"timestamp": "2026-03-23T08:54:48.175790Z", "level": "error", "event": "Training error: Failed to format ChatML dataset: One of the subprocesses has abruptly died during map operation.To debug the error, disable multiprocessing."}
instead of ModuleNotFoundError ..
from the first dude
wait
in which tab are you getting this?
because this is what might be the actual problem: AttributeError: Can't get attribute 'unsloth_push_to_hub' on <module 'unsloth.save' from
and was this lora training or full trainig
@main raptor can you show me what this gives you C:\Users\andi\.unsloth\studio\.venv\bin\pip show datasets
terminal
in which tab on the studio application
it was during qlora training*
let me grab the error in studio
i apologize for not being the most helpful here roland lol
and the attribute error follows this in powershell
have you seen this roland
the folder .venv does not exist
you probably updated
so we removed .venv
we now using unsloth_studio instead
C:\Users\andi\.unsloth\studio\unsloth_studio\Scripts\pip show datasets
Version: 4.3.0
Summary: HuggingFace community-driven open-source library of datasets
Home-page: https://github.com/huggingface/datasets
Author: HuggingFace Inc.
Author-email: [email protected]
License: Apache 2.0
Location: C:\Users\andi\.unsloth\studio\unsloth_studio\Lib\site-packages
Requires: dill, filelock, fsspec, httpx, huggingface-hub, multiprocess, numpy, packaging, pandas, pyarrow, pyyaml, requests, tqdm, xxhash
Required-by: evaluate, trl, unsloth, unsloth_zoo```
Name: numpy
Version: 2.4.3
Summary: Fundamental package for array computing in Python
Home-page: https://numpy.org
Author: Travis E. Oliphant et al.
Author-email:
License-Expression: BSD-3-Clause AND 0BSD AND MIT AND Zlib AND CC0-1.0
Location: C:\Users\andi\.unsloth\studio\unsloth_studio\Lib\site-packages
Requires:
Required-by: accelerate, bitsandbytes, contourpy, data-designer-config, datasets, descript-audio-codec, descript-audiotools, diffusers, einx, evaluate, langid, librosa, matplotlib, numba, onnxruntime, openai-whisper, pandas, peft, pyloudnorm, pymupdf-layout, pystoi, scikit-learn, scipy, snac, soundfile, soxr, tensorboard, torch-stoi, torchvision, transformers, transformers-cfg, unsloth, unsloth_zoo, xformers```
but i didnt tested with the newest version again (yet)
ok can you try installing the new version first . let's see if that solves your problem
