+ conda run -n finetuning --no-capture-output python mistral7b_train.py
Unsloth: Will load unsloth/mistral-7b-instruct-v0.3-bnb-4bit as a legacy tokenizer.
Using the latest cached version of the dataset since medmcqa couldn't be found on the Hugging Face Hub
Found the latest cached dataset configuration 'default' at /leonardo/home/userexternal/mpfister/.cache/huggingface/datasets/medmcqa/default/0.0.0/91c6572c454088bf71b679ad90aa8dffcd0d5868 (last modified on Thu Aug 29 19:38:14 2024).
🦥 Unsloth: Will patch your computer to enable 2x faster free finetuning.
==((====))== Unsloth 2024.8: Fast Mistral patching. Transformers = 4.43.4.
Detected kernel version 4.18.0, which is below the recommended minimum of 5.5.0; this can cause the process to hang. It is recommended to upgrade the kernel to the minimum version or higher.
max_steps is given, it will override any value given in num_train_epochs
"-____-" Number of trainable parameters = 41,943,040
trainable params: 41,943,040 || all params: 7,289,966,592 || trainable%: 0.5754
100%|██████████| 100/100 [00:53<00:00, 1.86it/s]
/leonardo/home/userexternal/mpfister/.conda/envs/finetuning/lib/python3.11/site-packages/peft/utils/other.py:619: UserWarning: Unable to fetch remote file due to the following error (MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /unsloth/mistral-7b-instruct-v0.3-bnb-4bit/resolve/main/config.json (Caused by NewConnectionError('<urllib3.connection.HTTPSConnection object at 0x150ea3ee3d10>: Failed to establish a new connection: [Errno 101] Network is unreachable'))"), '(Request ID: f7b42e8c-3e90-4492-a1cc-1fbf976543d7)') - silently ignoring the lookup for the file config.json in unsloth/mistral-7b-instruct-v0.3-bnb-4bit.
warnings.warn(
/leonardo/home/userexternal/mpfister/.conda/envs/finetuning/lib/python3.11/site-packages/peft/utils/save_and_load.py:218: UserWarning: Could not find a config file in unsloth/mistral-7b-instruct-v0.3-bnb-4bit - will assume that the vocabulary was not modified.
Training speed: 14.9 samples/s (=14.9 samples/s/GPU)
Memory occupied on GPUs: 7.0 GB.
real 6m9.210s
user 3m22.330s
sys 0m20.955s
+ conda run -n finetuning --no-capture-output python mistral7b_test.py
Unsloth: Will load unsloth/mistral-7b-instruct-v0.3-bnb-4bit as a legacy tokenizer.
Unsloth 2024.8 patched 32 layers with 32 QKV layers, 32 O layers and 32 MLP layers.
Using the latest cached version of the dataset since medmcqa couldn't be found on the Hugging Face Hub
Found the latest cached dataset configuration 'default' at /leonardo/home/userexternal/mpfister/.cache/huggingface/datasets/medmcqa/default/0.0.0/91c6572c454088bf71b679ad90aa8dffcd0d5868 (last modified on Tue Sep 10 17:39:08 2024).
🦥 Unsloth: Will patch your computer to enable 2x faster free finetuning.
==((====))== Unsloth 2024.8: Fast Mistral patching. Transformers = 4.43.4.
