sparse_llama_7b_hf2_refined_web_70p_2024-03-28

thrunlab/sparse_llama_7b_hf2_refined_web_70p_2024-03-28

sparse_llama_7b_hf2_refined_web_70p_2024-03-28 is a text generation model from thrunlab. Use it when you need the model to write or continue text. It is set up for transformers.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

3% of all-time downloads

All-time downloads

421

Public

Parameters

6.7B

137 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors13.7 GB · 100%

At a glance

Task: Text Generation
Library: transformers
Model type: sparse_llama
Access: Public
Created: Mar 28, 2024
Updated: Mar 28, 2024
SHA: 14c2402f

Try a prompt

Base models

meta-llama/Llama-2-7b-hf

Model card

From the Hugging Face model README

sparse_llama_7b_hf2_refined_web_70p_2024-03-28

This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the None dataset. It achieves the following results on the evaluation set:

Loss: 2.1370

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: 1
eval_batch_size: 4
seed: 0
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 4
total_train_batch_size: 16
total_eval_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
training_steps: 2600

Training results

Training Loss	Epoch	Step	Validation Loss
2.5973	0.0	25	2.4361
2.3245	0.01	50	2.4253
2.3668	0.01	75	2.4063
2.3521	0.02	100	2.3820
2.3119	0.02	125	2.3613
2.4636	0.02	150	2.3452
2.3372	0.03	175	2.3331
2.2927	0.03	200	2.3267
2.3831	0.04	225	2.3196
2.4013	0.04	250	2.3158
2.2576	0.04	275	2.3109
2.2803	0.05	300	2.3104
2.3531	0.05	325	2.3070
2.0961	0.06	350	2.3066
2.2809	0.06	375	2.3042
2.4516	0.06	400	2.3034
2.3927	0.07	425	2.3027
2.3188	0.07	450	2.3030
2.2165	0.08	475	2.3028
2.2814	0.08	500	2.3003
2.3224	0.08	525	2.3005
2.2267	0.09	550	2.2991
2.3238	0.09	575	2.2976
2.1845	0.1	600	2.2992
2.2896	0.1	625	2.2994
2.334	0.1	650	2.2973
2.3298	0.11	675	2.2972
2.2847	0.11	700	2.2950
2.2589	0.12	725	2.2966
2.1715	0.12	750	2.2949
2.3554	0.12	775	2.2940
2.3217	0.13	800	2.2970
2.3374	0.13	825	2.2940
2.2951	0.14	850	2.2933
2.1613	0.14	875	2.2934
2.278	0.14	900	2.2928
2.1964	0.15	925	2.2930
2.3489	0.15	950	2.2926
2.3561	0.16	975	2.2919
2.3854	0.16	1000	2.2907
2.3849	0.16	1025	2.2931
2.2953	0.17	1050	2.2917
2.3519	0.17	1075	2.2901
2.2702	0.18	1100	2.2927
2.2834	0.18	1125	2.2888
2.4015	0.18	1150	2.2925
2.3967	0.19	1175	2.2915
2.1978	0.19	1200	2.2892
2.2758	0.2	1225	2.2886
2.2774	0.2	1250	2.2905
2.3803	0.2	1275	2.2930
2.3233	0.21	1300	2.2909
2.2431	0.21	1325	2.2908
2.2042	0.22	1350	2.2885
2.3585	0.22	1375	2.2885
2.2779	0.22	1400	2.2906
2.2119	0.23	1425	2.2883
2.2257	0.23	1450	2.2876
2.2383	0.24	1475	2.2892
2.2166	0.24	1500	2.2899
2.1817	0.24	1525	2.2872
2.3113	0.25	1550	2.2880
2.3212	0.25	1575	2.2876
2.3027	0.26	1600	2.2838
2.2737	0.26	1625	2.2888
2.3181	0.26	1650	2.2865
2.2759	0.27	1675	2.2871
2.3179	0.27	1700	2.2858
2.3111	0.28	1725	2.2862
2.2699	0.28	1750	2.2872
2.3006	0.28	1775	2.2859
2.1913	0.29	1800	2.2849
2.3133	0.29	1825	2.2852
2.286	0.3	1850	2.2849
2.3403	0.3	1875	2.2860
2.2713	0.3	1900	2.2843
2.3492	0.31	1925	2.2840
2.2747	0.31	1950	2.2840
2.2783	0.32	1975	2.2841
2.1854	0.32	2000	2.2836
2.2544	0.32	2025	2.2826
2.2955	0.33	2050	2.2837
2.2053	0.33	2075	2.2849
2.2848	0.34	2100	2.2841
2.4168	0.34	2125	2.2843
2.268	0.34	2150	2.2814
2.2045	0.35	2175	2.2841
2.2717	0.35	2200	2.2825
2.1958	0.36	2225	2.2827
2.1635	0.36	2250	2.2867
2.3043	0.36	2275	2.2820
2.3106	0.37	2300	2.2845
2.1992	0.37	2325	2.2829
2.2189	0.38	2350	2.2817
2.3096	0.38	2375	2.2830
2.2803	0.38	2400	2.2839
2.1752	0.39	2425	2.2825
2.1324	0.39	2450	2.2834
2.207	0.4	2475	2.2846
2.3369	0.4	2500	2.2844
2.3512	0.4	2525	2.2853
2.1432	0.41	2550	2.2861
2.1743	0.41	2575	2.2832
2.2696	0.42	2600	2.2832

Framework versions

Transformers 4.40.0.dev0
Pytorch 2.1.1+cu121
Datasets 2.15.0
Tokenizers 0.15.2

Task: Text Generation
Library: transformers
Type: sparse_llama
Created: Mar 28, 2024
Updated: Mar 28, 2024