alexandersdata
/

gemma-4-E4B-hebrew-first

Generated from Trainer

Model card Files Files and versions

gemma-4-E4B-hebrew-first

840 MB

Ctrl+K

Ctrl+K

1 contributor

History: 9 commits

ssdata analysis

Model save

e938a9d verified 6 months ago

.gitattributes

1.57 kB
Training in progress, step 500 6 months ago
README.md

1.45 kB
Model save 6 months ago
adapter_config.json

10.2 kB
Training in progress, step 500 6 months ago
adapter_model.safetensors

808 MB
xet

Training in progress, step 500 6 months ago
chat_template.jinja

16.8 kB
Training in progress, step 500 6 months ago
tokenizer.json

32.2 MB
xet

Training in progress, step 500 6 months ago
tokenizer_config.json

2.74 kB
Training in progress, step 500 6 months ago
train.py

6.72 kB
Fix: add assistant_only_loss=False to prevent all labels being masked to -100 6 months ago
training_args.bin
Detected Pickle imports (10)
- "transformers.trainer_utils.SchedulerType",
- "transformers.trainer_pt_utils.AcceleratorConfig",
- "accelerate.state.PartialState",
- "transformers.trainer_utils.HubStrategy",
- "transformers.training_args.OptimizerNames",
- "transformers.trainer_utils.SaveStrategy",
- "trl.trainer.sft_config.SFTConfig",
- "torch.device",
- "accelerate.utils.dataclasses.DistributedType",
- "transformers.trainer_utils.IntervalStrategy"
How to fix it?
5.78 kB
xet

Training in progress, step 500 6 months ago