Quantization made by Richard Erkhov.

llama3.2-3B-Wolof - GGUF

Model creator: https://huggingface.co/Hawoly18/
Original model: https://huggingface.co/Hawoly18/llama3.2-3B-Wolof/

Name	Quant method	Size
llama3.2-3B-Wolof.Q2_K.gguf	Q2_K	1.27GB
llama3.2-3B-Wolof.IQ3_XS.gguf	IQ3_XS	1.37GB
llama3.2-3B-Wolof.IQ3_S.gguf	IQ3_S	1.43GB
llama3.2-3B-Wolof.Q3_K_S.gguf	Q3_K_S	1.43GB
llama3.2-3B-Wolof.IQ3_M.gguf	IQ3_M	1.49GB
llama3.2-3B-Wolof.Q3_K.gguf	Q3_K	1.57GB
llama3.2-3B-Wolof.Q3_K_M.gguf	Q3_K_M	1.57GB
llama3.2-3B-Wolof.Q3_K_L.gguf	Q3_K_L	1.69GB
llama3.2-3B-Wolof.IQ4_XS.gguf	IQ4_XS	1.71GB
llama3.2-3B-Wolof.Q4_0.gguf	Q4_0	1.78GB
llama3.2-3B-Wolof.IQ4_NL.gguf	IQ4_NL	1.79GB
llama3.2-3B-Wolof.Q4_K_S.gguf	Q4_K_S	1.79GB
llama3.2-3B-Wolof.Q4_K.gguf	Q4_K	1.88GB
llama3.2-3B-Wolof.Q4_K_M.gguf	Q4_K_M	1.88GB
llama3.2-3B-Wolof.Q4_1.gguf	Q4_1	1.95GB
llama3.2-3B-Wolof.Q5_0.gguf	Q5_0	2.11GB
llama3.2-3B-Wolof.Q5_K_S.gguf	Q5_K_S	2.11GB
llama3.2-3B-Wolof.Q5_K.gguf	Q5_K	2.16GB
llama3.2-3B-Wolof.Q5_K_M.gguf	Q5_K_M	2.16GB
llama3.2-3B-Wolof.Q5_1.gguf	Q5_1	2.27GB
llama3.2-3B-Wolof.Q6_K.gguf	Q6_K	2.46GB
llama3.2-3B-Wolof.Q8_0.gguf	Q8_0	3.18GB

Original model description:

library_name: transformers license: llama3.2 base_model: meta-llama/Llama-3.2-3B tags: - trl - sft - generated_from_trainer model-index: - name: outputs results: []

outputs

This model is a fine-tuned version of meta-llama/Llama-3.2-3B on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.6534

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 1
eval_batch_size: 8
seed: 3407
gradient_accumulation_steps: 4
total_train_batch_size: 4
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 5
training_steps: 100

Training results

Training Loss	Epoch	Step	Validation Loss
2.4702	0.0556	25	2.5017
2.1788	0.1111	50	2.0390
1.8193	0.1667	75	1.8122
1.5859	0.2222	100	1.6534

Framework versions

Transformers 4.45.2
Pytorch 2.1.0+cu118
Datasets 3.0.1
Tokenizers 0.20.1

Additional thanks to @nicoboss for giving me access to his private supercomputer, enabling me to provide many more quants, at much higher speed, than I would otherwise be able to.