Hugging Face's logo

RefalMachine
/
llm_test_unigram

Text Generation
Transformers
PyTorch
llama
Generated from Trainer
text-generation-inference
Model card Files Community
llm_test_unigram / README.md
RefalMachine's picture
RefalMachine
load model
50751bc
preview code
|
raw
history blame contribute delete
9.39 kB
metadata
base_model: llama7b_rulm_spm_unigram_10_mean_init_tie_12_10_23
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: llama7b_rulm_spm_unigram_10_mean_init_tie_rulm_small_1e_12_10_23
    results: []

llama7b_rulm_spm_unigram_10_mean_init_tie_rulm_small_1e_12_10_23

This model is a fine-tuned version of llama7b_rulm_spm_unigram_10_mean_init_tie_12_10_23 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 2.9526
  • Accuracy: 0.4388

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: 12
  • eval_batch_size: 12
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 10
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 240
  • total_eval_batch_size: 120
  • 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
3.5018 0.01 1000 3.5095 0.3768
3.3375 0.02 2000 3.3603 0.3918
3.2694 0.02 3000 3.2995 0.3982
3.236 0.03 4000 3.2626 0.4022
3.2131 0.04 5000 3.2350 0.4048
3.1934 0.05 6000 3.2147 0.4073
3.177 0.06 7000 3.2006 0.4089
3.1653 0.07 8000 3.1890 0.4097
3.1548 0.07 9000 3.1779 0.4109
3.152 0.08 10000 3.1699 0.4119
3.1416 0.09 11000 3.1622 0.4130
3.1387 0.1 12000 3.1561 0.4139
3.1192 0.11 13000 3.1495 0.4143
3.1221 0.12 14000 3.1431 0.4150
3.1136 0.12 15000 3.1391 0.4155
3.106 0.13 16000 3.1348 0.4160
3.1023 0.14 17000 3.1312 0.4165
3.1062 0.15 18000 3.1265 0.4172
3.1007 0.16 19000 3.1230 0.4177
3.0979 0.16 20000 3.1201 0.4178
3.0897 0.17 21000 3.1168 0.4179
3.0863 0.18 22000 3.1128 0.4189
3.0898 0.19 23000 3.1097 0.4191
3.0825 0.2 24000 3.1074 0.4191
3.0808 0.21 25000 3.1037 0.4200
3.0774 0.21 26000 3.1032 0.4197
3.0652 0.22 27000 3.0980 0.4202
3.0693 0.23 28000 3.0968 0.4207
3.0665 0.24 29000 3.0944 0.4209
3.0657 0.25 30000 3.0920 0.4210
3.0608 0.26 31000 3.0911 0.4213
3.0647 0.26 32000 3.0896 0.4213
3.0604 0.27 33000 3.0861 0.4217
3.0577 0.28 34000 3.0845 0.4221
3.0606 0.29 35000 3.0814 0.4220
3.0515 0.3 36000 3.0801 0.4227
3.0527 0.31 37000 3.0772 0.4225
3.0507 0.31 38000 3.0758 0.4228
3.0433 0.32 39000 3.0739 0.4234
3.0546 0.33 40000 3.0717 0.4234
3.0484 0.34 41000 3.0697 0.4236
3.0441 0.35 42000 3.0694 0.4236
3.0292 0.35 43000 3.0662 0.4242
3.0384 0.36 44000 3.0643 0.4244
3.0367 0.37 45000 3.0629 0.4240
3.0337 0.38 46000 3.0622 0.4246
3.0385 0.39 47000 3.0599 0.4245
3.0319 0.4 48000 3.0574 0.4250
3.0255 0.4 49000 3.0573 0.4249
3.021 0.41 50000 3.0557 0.4253
3.0305 0.42 51000 3.0530 0.4254
3.0248 0.43 52000 3.0528 0.4257
3.0269 0.44 53000 3.0495 0.4261
3.0136 0.45 54000 3.0488 0.4259
3.0156 0.45 55000 3.0468 0.4262
3.022 0.46 56000 3.0454 0.4268
3.0193 0.47 57000 3.0442 0.4269
3.0222 0.48 58000 3.0417 0.4270
3.0111 0.49 59000 3.0393 0.4276
3.0148 0.49 60000 3.0384 0.4273
3.0077 0.5 61000 3.0364 0.4276
3.0167 0.51 62000 3.0358 0.4276
3.0049 0.52 63000 3.0343 0.4280
3.016 0.53 64000 3.0322 0.4281
3.0103 0.54 65000 3.0297 0.4285
3.0066 0.54 66000 3.0290 0.4284
2.9958 0.55 67000 3.0281 0.4285
3.0062 0.56 68000 3.0266 0.4288
2.9985 0.57 69000 3.0245 0.4289
3.0031 0.58 70000 3.0224 0.4292
2.9894 0.59 71000 3.0214 0.4295
2.9929 0.59 72000 3.0193 0.4296
2.9904 0.6 73000 3.0176 0.4296
2.9989 0.61 74000 3.0171 0.4301
2.9959 0.62 75000 3.0153 0.4301
2.9847 0.63 76000 3.0142 0.4306
2.9892 0.63 77000 3.0127 0.4308
2.9924 0.64 78000 3.0110 0.4310
2.991 0.65 79000 3.0096 0.4312
2.9824 0.66 80000 3.0080 0.4311
2.9879 0.67 81000 3.0060 0.4315
2.9764 0.68 82000 3.0042 0.4321
2.9827 0.68 83000 3.0030 0.4315
2.9769 0.69 84000 3.0012 0.4324
2.9788 0.7 85000 3.0002 0.4322
2.9734 0.71 86000 2.9987 0.4325
2.9769 0.72 87000 2.9975 0.4328
2.9676 0.73 88000 2.9959 0.4326
2.9677 0.73 89000 2.9943 0.4330
2.9739 0.74 90000 2.9933 0.4330
2.9691 0.75 91000 2.9914 0.4334
2.969 0.76 92000 2.9901 0.4336
2.9602 0.77 93000 2.9889 0.4337
2.965 0.78 94000 2.9872 0.4339
2.9627 0.78 95000 2.9853 0.4341
2.9542 0.79 96000 2.9844 0.4340
2.9552 0.8 97000 2.9822 0.4344
2.9576 0.81 98000 2.9812 0.4347
2.9579 0.82 99000 2.9802 0.4348
2.9508 0.82 100000 2.9784 0.4349
2.9551 0.83 101000 2.9771 0.4353
2.9535 0.84 102000 2.9759 0.4357
2.9479 0.85 103000 2.9743 0.4357
2.9542 0.86 104000 2.9732 0.4359
2.9481 0.87 105000 2.9715 0.4360
2.941 0.87 106000 2.9697 0.4362
2.9435 0.88 107000 2.9684 0.4365
2.9403 0.89 108000 2.9674 0.4368
2.9453 0.9 109000 2.9661 0.4367
2.9396 0.91 110000 2.9644 0.4372
2.9375 0.92 111000 2.9633 0.4372
2.9284 0.92 112000 2.9621 0.4374
2.9418 0.93 113000 2.9606 0.4376
2.934 0.94 114000 2.9594 0.4377
2.9374 0.95 115000 2.9583 0.4380
2.9302 0.96 116000 2.9569 0.4382
2.9273 0.96 117000 2.9560 0.4382
2.9338 0.97 118000 2.9548 0.4384
2.9304 0.98 119000 2.9539 0.4385
2.9361 0.99 120000 2.9531 0.4385
2.927 1.0 121000 2.9526 0.4387

Framework versions

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