RichardErkhov/interview-eval_-_zephyr-7b-math-test-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

zephyr-7b-math-test - GGUF

• Model creator: https://huggingface.co/interview-eval/
• Original model: https://huggingface.co/interview-eval/zephyr-7b-math-test/

Name	Quant method	Size
zephyr-7b-math-test.Q2_K.gguf	Q2_K	2.53GB
zephyr-7b-math-test.IQ3_XS.gguf	IQ3_XS	2.81GB
zephyr-7b-math-test.IQ3_S.gguf	IQ3_S	2.96GB
zephyr-7b-math-test.Q3_K_S.gguf	Q3_K_S	2.95GB
zephyr-7b-math-test.IQ3_M.gguf	IQ3_M	3.06GB
zephyr-7b-math-test.Q3_K.gguf	Q3_K	3.28GB
zephyr-7b-math-test.Q3_K_M.gguf	Q3_K_M	3.28GB
zephyr-7b-math-test.Q3_K_L.gguf	Q3_K_L	3.56GB
zephyr-7b-math-test.IQ4_XS.gguf	IQ4_XS	3.67GB
zephyr-7b-math-test.Q4_0.gguf	Q4_0	3.83GB
zephyr-7b-math-test.IQ4_NL.gguf	IQ4_NL	3.87GB
zephyr-7b-math-test.Q4_K_S.gguf	Q4_K_S	3.86GB
zephyr-7b-math-test.Q4_K.gguf	Q4_K	4.07GB
zephyr-7b-math-test.Q4_K_M.gguf	Q4_K_M	4.07GB
zephyr-7b-math-test.Q4_1.gguf	Q4_1	4.24GB
zephyr-7b-math-test.Q5_0.gguf	Q5_0	4.65GB
zephyr-7b-math-test.Q5_K_S.gguf	Q5_K_S	4.65GB
zephyr-7b-math-test.Q5_K.gguf	Q5_K	4.78GB
zephyr-7b-math-test.Q5_K_M.gguf	Q5_K_M	4.78GB
zephyr-7b-math-test.Q5_1.gguf	Q5_1	5.07GB
zephyr-7b-math-test.Q6_K.gguf	Q6_K	5.53GB
zephyr-7b-math-test.Q8_0.gguf	Q8_0	7.17GB

Original model description:

library_name: transformers license: apache-2.0 base_model: alignment-handbook/zephyr-7b-sft-full tags: - alignment-handbook - trl - sft - generated_from_trainer - trl - sft - generated_from_trainer datasets: - EunsuKim/MATH model-index: - name: zephyr-7b-math-test results: []

zephyr-7b-math-test

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the EunsuKim/MATH dataset. It achieves the following results on the evaluation set:

• Loss: 0.0178

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 64
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss
0.8684	1.0	5	0.7914
0.7031	2.0	10	0.5349
0.4888	3.0	15	0.3412
0.302	4.0	20	0.1902
0.1643	5.0	25	0.0890
0.0778	6.0	30	0.0416
0.0404	7.0	35	0.0280
0.0279	8.0	40	0.0219
0.0214	9.0	45	0.0186
0.0183	10.0	50	0.0178

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu124
• Datasets 2.21.0
• Tokenizers 0.19.1