RefalMachine/llm_model_m_unigram_hm_1e_v0.1

mistral7b_darulm_unigram_1e_20_10_23

This model is a fine-tuned version of mistral_rulm_unigram_init_20_10_23 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.7017
• Accuracy: 0.4706

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: 0.0003
• train_batch_size: 6
• eval_batch_size: 6
• seed: 42
• distributed_type: multi-GPU
• num_devices: 10
• gradient_accumulation_steps: 4
• total_train_batch_size: 240
• total_eval_batch_size: 60
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.1631	0.01	1000	3.1804	0.4182
3.1124	0.02	2000	3.1065	0.4272
3.0757	0.03	3000	3.0894	0.4288
3.0488	0.04	4000	3.0630	0.4319
3.0403	0.05	5000	3.0423	0.4336
3.0172	0.05	6000	3.0384	0.4343
3.0102	0.06	7000	3.0267	0.4360
2.9888	0.07	8000	3.0190	0.4361
3.0024	0.08	9000	3.0040	0.4385
2.9948	0.09	10000	3.0058	0.4378
2.9774	0.1	11000	2.9962	0.4389
2.9818	0.11	12000	2.9964	0.4390
2.9771	0.12	13000	2.9913	0.4396
2.9786	0.13	14000	2.9915	0.4391
2.9866	0.14	15000	2.9924	0.4394
2.9751	0.14	16000	2.9918	0.4389
2.9702	0.15	17000	2.9926	0.4393
2.9695	0.16	18000	2.9816	0.4401
2.9615	0.17	19000	2.9826	0.4402
2.9609	0.18	20000	2.9791	0.4406
2.9607	0.19	21000	2.9684	0.4416
2.9533	0.2	22000	2.9677	0.4422
2.9513	0.21	23000	2.9676	0.4421
2.9563	0.22	24000	2.9610	0.4429
2.9466	0.23	25000	2.9627	0.4424
2.9431	0.24	26000	2.9590	0.4424
2.9412	0.24	27000	2.9525	0.4436
2.9299	0.25	28000	2.9504	0.4435
2.9332	0.26	29000	2.9486	0.4435
2.9255	0.27	30000	2.9425	0.4442
2.9242	0.28	31000	2.9459	0.4434
2.9242	0.29	32000	2.9378	0.4445
2.9267	0.3	33000	2.9316	0.4453
2.9151	0.31	34000	2.9315	0.4454
2.9105	0.32	35000	2.9286	0.4456
2.9053	0.33	36000	2.9242	0.4457
2.9023	0.33	37000	2.9195	0.4466
2.8946	0.34	38000	2.9177	0.4468
2.9037	0.35	39000	2.9147	0.4470
2.8893	0.36	40000	2.9130	0.4468
2.8891	0.37	41000	2.9055	0.4481
2.8851	0.38	42000	2.9017	0.4485
2.8909	0.39	43000	2.9011	0.4483
2.896	0.4	44000	2.9061	0.4479
2.8918	0.41	45000	2.9043	0.4479
2.8847	0.42	46000	2.8954	0.4490
2.8749	0.42	47000	2.8912	0.4494
2.8832	0.43	48000	2.8912	0.4496
2.8745	0.44	49000	2.8853	0.4500
2.8717	0.45	50000	2.8834	0.4502
2.8659	0.46	51000	2.8831	0.4503
2.865	0.47	52000	2.8784	0.4505
2.8575	0.48	53000	2.8763	0.4508
2.8571	0.49	54000	2.8741	0.4513
2.8554	0.5	55000	2.8704	0.4514
2.8526	0.51	56000	2.8669	0.4519
2.8521	0.52	57000	2.8618	0.4525
2.8398	0.52	58000	2.8600	0.4522
2.8398	0.53	59000	2.8576	0.4528
2.837	0.54	60000	2.8536	0.4528
2.837	0.55	61000	2.8519	0.4535
2.8427	0.56	62000	2.8493	0.4536
2.8365	0.57	63000	2.8468	0.4541
2.8327	0.58	64000	2.8447	0.4539
2.8289	0.59	65000	2.8388	0.4546
2.8166	0.6	66000	2.8346	0.4547
2.8171	0.61	67000	2.8294	0.4558
2.8184	0.61	68000	2.8269	0.4556
2.8102	0.62	69000	2.8243	0.4563
2.8153	0.63	70000	2.8211	0.4564
2.8035	0.64	71000	2.8185	0.4569
2.8042	0.65	72000	2.8206	0.4569
2.7984	0.66	73000	2.8138	0.4574
2.7883	0.67	74000	2.8112	0.4574
2.7962	0.68	75000	2.8056	0.4584
2.7937	0.69	76000	2.8068	0.4582
2.7853	0.7	77000	2.8011	0.4588
2.7798	0.71	78000	2.7954	0.4597
2.7851	0.71	79000	2.7913	0.4598
2.7831	0.72	80000	2.7897	0.4600
2.7773	0.73	81000	2.7862	0.4603
2.7688	0.74	82000	2.7836	0.4609
2.7658	0.75	83000	2.7798	0.4610
2.7622	0.76	84000	2.7815	0.4612
2.7691	0.77	85000	2.7783	0.4612
2.7579	0.78	86000	2.7712	0.4619
2.7614	0.79	87000	2.7673	0.4625
2.7592	0.8	88000	2.7691	0.4623
2.7551	0.8	89000	2.7607	0.4634
2.7397	0.81	90000	2.7579	0.4637
2.7357	0.82	91000	2.7580	0.4636
2.7452	0.83	92000	2.7517	0.4643
2.7418	0.84	93000	2.7533	0.4641
2.7379	0.85	94000	2.7481	0.4647
2.7308	0.86	95000	2.7460	0.4654
2.727	0.87	96000	2.7408	0.4655
2.7282	0.88	97000	2.7351	0.4664
2.7133	0.89	98000	2.7301	0.4669
2.7136	0.9	99000	2.7251	0.4673
2.7108	0.9	100000	2.7208	0.4679
2.7051	0.91	101000	2.7192	0.4681
2.7013	0.92	102000	2.7151	0.4687
2.6996	0.93	103000	2.7129	0.4689
2.6898	0.94	104000	2.7084	0.4694
2.688	0.95	105000	2.7053	0.4697
2.6855	0.96	106000	2.7018	0.4701
2.6852	0.97	107000	2.6989	0.4705
2.689	0.98	108000	2.6982	0.4705
2.6868	0.99	109000	2.6994	0.4707
2.6901	0.99	110000	2.7006	0.4707

Framework versions

• Transformers 4.34.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1