#Finetuning Ministral-3-3B-Instruct-2512

9 messages · Page 1 of 1 (latest)

sage chasm
#

Not sure why the original message was deleted, but alright!

#

When trying to load the model:

model, tokenizer = FastVisionModel.from_pretrained(
    model_name = "unsloth/Ministral-3-3B-Instruct-2512",
    max_seq_length = 2048,             # Context length - can be longer, but uses more memory
    load_in_4bit = False,              # 4bit uses much less memory
    full_finetuning = True,            # We have full finetuning now!
    #token = "hf_...",
)

...I get this:

RuntimeError: Unsloth: No config file found - are you sure the `model_name` is correct?
If you're using a model on your local device, confirm if the folder location exists.
If you're using a HuggingFace online model, check if it exists.
#

Setting the token to my own doesn't fix the issue

#

Here is how I set up my project:

uv venv --no-project --relocatable --python=3.13
.venv\Scripts\activate
uv pip install unsloth transformers trl
uv pip install --force-reinstall torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu128
#

I'm very new to this all, so I'm sorry if I made obvious mistakes!

sage chasm
#

As a workaround, I got it to load when downloading the model first:

model_id = "unsloth/Ministral-3-3B-Instruct-2512"
model_path = f'./hf_hub/model--{model_id.replace("/", "--")}/'
snapshot_download(repo_id=model_id, local_dir=model_path)

model, tokenizer = FastVisionModel.from_pretrained(
    model_path,
    load_in_4bit = False,              # Use 4bit to reduce memory use. False for 16bit LoRA.
    max_seq_length = 2048,
    full_finetuning = True,
)
strong isle
#

Did you use the latest transformers?

#

you need to install from source ie pip install git+https://github.com/huggingface/transformers.git@bf3f0ae70d0e902efab4b8517fce88f6697636ce

finite dawn
#

Can confirm - installing from source works