This model is a fine-tuned version of Qwen/Qwen3-1.7B-Base on the 30b_a3b_math_95k_16_train dataset. It achieves the following results on the evaluation set:

Loss: 0.0101
Token Mean Mae: 25192073741.5337
Token Mean Rmse: 7814890.5454
Token Mean Seq Mean Mae: 3065597.9476
Token Mean Seq Mean Rmse: 51762.7254
Token Mean Relerr: 0.6543
Token Mean Seq Mean Relerr: 0.9140

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: 2e-05
train_batch_size: 4
eval_batch_size: 4
seed: 0
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 64
total_train_batch_size: 1024
total_eval_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 50
num_epochs: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss	Token Mean Mae	Token Mean Relerr	Token Mean Rmse	Token Mean Seq Mean Mae	Token Mean Seq Mean Relerr	Token Mean Seq Mean Rmse
0.8838	0.0337	50	0.0152	31962873332.2035	1.2905	9067506.1269	3893413.1969	1.5436	67383.2700
0.9150	0.0675	100	0.0130	29082752791.3909	1.0013	8487870.1899	3541616.5108	1.2401	62469.0928
0.7341	0.1012	150	0.0121	28461424534.9630	0.9021	8596978.0872	3464854.9661	1.0606	59611.2344
0.6296	0.1350	200	0.0114	26886060544.5233	0.8831	7827202.9967	3272673.0010	1.2330	59088.6549
0.6926	0.1687	250	0.0111	26591412583.9368	0.8925	7834076.2851	3236449.6441	1.1603	57334.6885
0.6139	0.2025	300	0.0108	26702937964.4749	0.7990	8092083.5720	3249971.2530	0.9607	56151.5083
0.5502	0.2362	350	0.0107	26309989800.4493	0.7767	7943355.2861	3202539.2777	1.0370	55998.6088
0.5278	0.2700	400	0.0111	27214946205.0021	0.7060	8392862.9583	3311882.0626	0.9070	56098.4729
0.4964	0.3037	450	0.0107	26375763016.5343	0.7337	8052422.8244	3210015.3622	0.9549	55505.9211
0.4826	0.3374	500	0.0107	26174723423.1948	0.7004	8016074.2561	3185610.6003	0.8731	54969.6635
0.4651	0.3712	550	0.0106	25923456973.7465	0.7817	7764392.2592	3155471.7348	1.0857	55274.2759
0.4350	0.4049	600	0.0107	26005477322.2059	0.7476	7835539.6734	3164525.7692	1.0798	54937.2762
0.4476	0.4387	650	0.0107	27094116153.3255	0.6869	8314936.4869	3297115.6696	0.8879	55315.1395
0.4649	0.4724	700	0.0103	26127305157.3229	0.7682	7911443.6463	3179186.8054	1.0535	54562.6628
0.4550	0.5062	750	0.0107	26642414381.7156	0.6877	8177778.9883	3241904.4287	0.9428	54600.9327
0.4041	0.5399	800	0.0109	26305712502.5043	0.5907	8205801.4943	3201468.6334	0.7266	53629.8888
0.4041	0.5737	850	0.0104	25481182626.3355	0.6144	7828342.2509	3101531.0826	0.8132	53606.8506
0.4068	0.6074	900	0.0103	25884806288.5230	0.7346	7809702.6584	3150509.8743	1.0223	54361.4305
0.3763	0.6411	950	0.0104	25686105330.9954	0.6695	7858742.2040	3125837.7557	0.8839	53570.9686
0.4063	0.6749	1000	0.0105	25726341541.9370	0.6561	7836993.6156	3131141.7075	0.8876	53564.3389
0.3715	0.7086	1050	0.0104	25577657636.7409	0.6572	7896683.5400	3112295.4313	0.8670	52933.0834
0.4101	0.7424	1100	0.0104	25697994539.2222	0.6821	7962045.3988	3128070.5844	0.9236	53051.6250
0.3296	0.7761	1150	0.0107	25976736045.6458	0.6162	8096339.3832	3161779.6041	0.7932	52897.0733
0.3780	0.8099	1200	0.0104	25615151552.0550	0.6382	7937833.8960	3117088.2811	0.8487	52774.9458
0.3768	0.8436	1250	0.0105	25941770840.4481	0.6784	7899501.1527	3156785.4411	0.9677	53674.3792
0.3788	0.8774	1300	0.0103	25405377328.5631	0.6109	7866119.5046	3091528.6967	0.7914	52509.0489
0.3696	0.9111	1350	0.0103	25532641757.3267	0.7108	7854749.4646	3107104.7411	0.9830	52819.0236
0.3682	0.9448	1400	0.0104	25367025567.7703	0.6413	7852615.6312	3086526.1212	0.8465	52453.1452
0.3389	0.9786	1450	0.0103	25576755953.8150	0.6636	7963191.7338	3112153.6540	0.8621	52445.6380
0.3944	1.0121	1500	0.0103	25320670727.8167	0.6504	7864982.3833	3081025.3964	0.8569	52112.9809
0.3105	1.0459	1550	0.0103	25486237905.7501	0.6173	7879228.6209	3100864.6937	0.8305	52328.9515
0.2800	1.0796	1600	0.0103	25488200028.0725	0.6656	7883183.3556	3101773.2576	0.9099	52448.8037
0.3174	1.1134	1650	0.0104	25861998661.0543	0.6113	7982692.4526	3146830.2875	0.8534	52835.8204
0.3140	1.1471	1700	0.0102	25521483105.0387	0.6678	7877753.0629	3105780.1889	0.9222	52315.0782
0.3094	1.1809	1750	0.0102	25419660857.8242	0.6686	7818757.9673	3093012.4016	0.9424	52418.1496
0.3283	1.2146	1800	0.0102	25102394546.4403	0.6894	7745066.5962	3055248.9459	0.9429	52037.3113
0.3099	1.2484	1850	0.0103	25604987186.0661	7944719.9374	3116075.6429	52485.8584	0.6500	0.9077
0.3335	1.2821	1900	0.0105	25588991772.1383	7976323.3992	3113838.4760	52169.5469	0.6070	0.8091
0.2886	1.3158	1950	0.0100	25196339839.8400	7774726.7876	3066369.9007	52171.6559	0.6687	0.9250
0.3390	1.3496	2000	0.0102	25348500890.8203	7850771.5124	3084686.9669	52099.1455	0.6378	0.8852
0.3311	1.3833	2050	0.0102	25315336130.1431	7864825.8588	3080836.5537	52034.2528	0.6473	0.8774
0.3267	1.4171	2100	0.0101	25085917619.3055	7637916.0089	3053308.5697	52557.2485	0.7049	1.0112
0.3251	1.4508	2150	0.0102	25324994805.8852	7870383.6193	3082062.2541	51898.1139	0.6509	0.8885
0.2910	1.4846	2200	0.0101	25160413607.9405	7790770.0594	3061667.9161	51900.3286	0.6562	0.9123
0.3220	1.5183	2250	0.0102	25440209435.1005	7869804.5901	3095594.9222	52125.3182	0.6508	0.9117
0.2914	1.5521	2300	0.0102	25518166445.5846	7879342.7310	3105138.3217	52255.2315	0.6712	0.9555
0.2871	1.5858	2350	0.0101	25144737131.4202	7778736.8958	3059855.0501	51879.8562	0.6620	0.9193
0.3406	1.6196	2400	0.0101	25245011524.3196	7818310.1363	3072268.1235	51888.5041	0.6657	0.9357
0.3289	1.6533	2450	0.0101	25212716686.9061	7774020.1217	3068524.1634	52036.5929	0.6817	0.9712
0.3103	1.6870	2500	0.0101	25199619583.6928	7802074.2305	3066695.1766	51900.3792	0.6777	0.9584
0.2776	1.7208	2550	0.0101	25249907548.2346	7854459.2565	3072602.8266	51775.4323	0.6385	0.8874
0.2726	1.7545	2600	0.0102	25253709515.6708	7843764.1642	3073097.8076	51811.0960	0.6519	0.9082
0.3006	1.7883	2650	0.0101	25252189159.9992	7837001.5254	3072967.6592	51812.2177	0.6503	0.9112
0.3168	1.8220	2700	0.0101	25202676950.5567	7824637.5506	3066863.8581	51768.4061	0.6489	0.9075
0.3199	1.8558	2750	0.0101	25252600129.8480	7835128.3652	3072959.5303	51824.7472	0.6557	0.9195
0.2877	1.8895	2800	0.0101	25183954796.2180	7813811.6609	3064627.4014	51755.0915	0.6560	0.9180
0.3383	1.9233	2850	0.0101	25204580360.5053	7822050.6876	3067122.0695	51757.7305	0.6527	0.9116
0.3070	1.9570	2900	0.0101	25188540475.4139	7813731.1015	3065199.7823	51756.3663	0.6544	0.9144
0.3022	1.9907	2950	0.0101	25189441970.3433	7814560.9700	3065276.1438	51757.7307	0.6544	0.9143

Framework versions

Transformers 5.0.0
Pytorch 2.10.0+cu128
Datasets 4.0.0
Tokenizers 0.22.2

namezz/lvm-math-0408-a-qwen3-30b-a3b-instruct-b-qwen3-1.7b-base

abs-bf16-30b-a3b-math-lr2e-5-g0.9998-l1.0-gpu4-bs4-ga64-ep2-wu50-cut15000

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

namezz/
lvm-math-0408-a-qwen3-30b-a3b-instruct-b-qwen3-1.7b-base