Ir para o conteúdo
ModelListadosExterno
Baixar

Iris Llama 3.2 3b Multihop Only K10

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the None dataset. It achieves the following results on the evaluation set: Loss: 0.8741 Model Preparation Time: 0.0212 Soft Mae: 0.0876 Soft...

Fontehuggingface.co/daryaZare/iris-llama-3.2-3b-multihop-only-k10/tree/d3d968ee1ec9363bbd3c05848c8e34f41a07d5b8
TipoOtherModelo basemeta-llama/Llama-3.2-3B-InstructVersãoAdapter modelBnota

Fonte do modelo

Tipo
Other
Tamanho do arquivo
92,8 MB
Modelo base
meta-llama/Llama-3.2-3B-Instruct
Versão
vAdapter model
Atualizado
15 de set. de 2026
Fonte
Hugging Face

Descrição da fonte

iris-llama-3.2-3b-multihop-only-k10

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.8741
  • Model Preparation Time: 0.0212
  • Soft Mae: 0.0876
  • Soft Brier: 0.0220
  • Student Prelevantmean: 0.3055
  • Teacher Prelevantmean: 0.3219
  • Bin F1: 0.8663
  • Cal Ece: 0.1361
  • Cal Brier: 0.0946
  • Cal Auroc: 0.9692
  • Info Ndcg@p8: 0.9724
  • Info Pairwiseacc: 0.7891
  • Num Questions: 20
  • Num Pairs: 253

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: 5e-05
  • train_batch_size: 1
  • eval_batch_size: 1
  • seed: 42
  • gradient_accumulation_steps: 16
  • total_train_batch_size: 16
  • optimizer: Use paged_adamw_8bit with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: constant
  • lr_scheduler_warmup_steps: 0.03
  • num_epochs: 6.0

Training results

Training LossEpochStepValidation LossModel Preparation TimeSoft MaeSoft BrierStudent PrelevantmeanTeacher PrelevantmeanBin F1Cal EceCal BrierCal AurocInfo Ndcg@p8Info PairwiseaccNum QuestionsNum Pairs
1.03470.2562381.05080.02120.14370.04710.22920.32190.61640.17000.15320.93540.95860.763720253
1.09670.5124761.01310.02120.13500.04020.27910.32190.69510.13830.13370.92710.94880.773620253
0.86730.76861140.87040.02120.12190.03250.26980.32190.71340.14780.12320.95920.94930.792120253
0.90891.02021520.93280.02120.12720.03550.34500.32190.80200.10070.11180.94240.96080.791320253
0.77391.27641900.89360.02120.12890.03680.36320.32190.83840.10610.11120.93690.95510.775020253
0.70581.53272280.92100.02120.12040.03290.35890.32190.81820.07150.10150.95000.94960.792620253
0.65211.78892660.78960.02120.09960.02550.29090.32190.80230.12240.10440.96450.95990.800420253
0.71862.04053040.86650.02120.10720.03010.27340.32190.80230.13890.11200.96180.96450.790220253
0.53982.29673420.86960.02120.10640.03160.25700.32190.77650.15710.11840.96200.96460.786520253
0.48822.55293800.85380.02120.10350.02800.26780.32190.79070.14500.11080.96660.96550.800820253
0.43462.80914180.83200.02120.10090.02570.32890.32190.84690.10120.10090.96070.96470.806320253
0.37173.06074561.03670.02120.10260.02870.26650.32190.76650.15080.10790.97370.96870.801020253
0.35103.31694940.91980.02120.09960.02740.31910.32190.84780.12710.10000.96050.96970.789520253
0.47003.57315320.95010.02120.11130.03200.25950.32190.72150.15200.12220.96180.96510.790920253
0.36643.82935701.01330.02120.09330.02440.31510.32190.84820.10640.10050.96240.98090.800620253
0.33254.08096080.96610.02120.10900.02870.27550.32190.75740.14170.11350.96450.96690.778420253
0.22554.33716461.08840.02120.09630.02410.30020.32190.85710.12960.09990.96680.97430.796320253
0.33304.59336841.11700.02120.11340.03210.24880.32190.69620.16360.12190.96660.96190.788420253
0.23804.84967221.02460.02120.10660.02860.29160.32190.81770.12520.10970.95750.96910.789720253
0.23785.10117600.86400.02120.09170.02410.29080.32190.84090.13220.09920.97110.97010.804620253
0.26545.35747981.12000.02120.10340.03030.27340.32190.78820.14470.10950.96980.96660.797220253
0.24905.61368361.11570.02120.09940.02880.30790.32190.83060.10020.10140.95650.96750.791120253
0.30385.86988740.91860.02120.08770.02380.28880.32190.83980.12720.09750.97170.96820.799920253
0.17936.08940.87410.02120.08760.02200.30550.32190.86630.13610.09460.96920.97240.789120253

Framework versions

  • PEFT 0.19.1
  • Transformers 5.14.1
  • Pytorch 2.5.1+cu124
  • Datasets 5.0.0
  • Tokenizers 0.22.2

Fontes

1 fonte
Hugging FacePrimárioVersão d3d968ee1ec9363bbd3c05848c8e34f41a07d5b8
huggingface.co/daryaZare/iris-llama-3.2-3b-multihop-only-k10/tree/d3d968ee1ec9363bbd3c05848c8e34f41a07d5b8

Verificado 8 de ago.

Artefatos de modelo

13 artefatos

adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 530cf8ce1871…dea2 · Hugging Face

Baixar

checkpoint-150/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 84ae0a462db5…79f0 · Hugging Face

Baixar

checkpoint-225/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 8e42c15c116f…98c6 · Hugging Face

Baixar

checkpoint-300/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 ac5975e3df69…d94b · Hugging Face

Baixar

checkpoint-375/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 739d0aa57551…ca4b · Hugging Face

Baixar

checkpoint-450/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 b592134b5679…bbbe · Hugging Face

Baixar

checkpoint-525/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 33c2a6d32c50…ab0f · Hugging Face

Baixar

checkpoint-600/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 ed0335a27710…dffb · Hugging Face

Baixar

checkpoint-675/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 a621dae80b23…aa4d · Hugging Face

Baixar

checkpoint-75/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 10bf7352b20a…dc43 · Hugging Face

Baixar

checkpoint-750/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 3319f5bd8623…924b · Hugging Face

Baixar

checkpoint-825/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 231fc127c761…402e · Hugging Face

Baixar

checkpoint-894/adapter_model.safetensors

safetensors · 92,8 MB · SHA-256 530cf8ce1871…dea2 · Hugging Face

Baixar

Trechos de fonte

2 trechos
Readme8 de ago.huggingface.co/daryaZare/iris-llama-3.2-3b-multihop-only-k10

--- library_name: peft license: llama3.2 base_model: meta-llama/Llama-3.2-3B-Instruct tags: - base_model:adapter:meta-llama/Llama-3.2-3B-Instruct - lora - transformers pipeline_tag: text-generation model-index: - name: iris-llama-3.2-3b-multihop-only-k10 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # iris-llama-3.2-3b-multihop-only-k10 This model is a fine-tuned version of [meta-llama/Llama-3.2-3B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-3B-Instruct) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.8741 - Model Preparation Time: 0.0212 - Soft Mae: 0.0876 - Soft Brier: 0.0220 - Student Prelevantmean: 0.3055 - Teacher Prelevantmean: 0.3219 - Bin F1: 0.8663 - Cal Ece: 0.1361 - Cal Brier: 0.0946 - Cal Auroc: 0.9692 - Info Ndcg@p8: 0.9724 - Info Pairwiseacc: 0.7891 - Num Questions: 20 - Num Pairs: 253 ## 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: 5e-05 - train_batch_size: 1 - eval_batch_size: 1 - seed: 42 - gradient_accumulation_steps: 16 - total_train_batch_size: 16 - optimizer: Use paged_adamw_8bit with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments - lr_scheduler_type: constant - lr_scheduler_warmup_steps: 0.03 - num_epochs: 6.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Soft Mae | Soft Brier | Student Prelevantmean | Teacher Prelevantmean | Bin F1 | Cal Ece | Cal Brier | Cal Auroc | Info Ndcg@p8 | Info Pairwiseacc | Num Questions | Num Pairs | |:-------------:|:------:|:----:|:---------------:|:----------------------:|:--------:|:----------:|:---------------------:|:---------------------:|:------:|:-------:|:---------:|:---------:|:------------:|:----------------:|:-------------:|:---------:| | 1.0347 | 0.2562 | 38 | 1.0508 | 0.0212 | 0.1437 | 0.0471 | 0.2292 | 0.3219 | 0.6164 | 0.1700 | 0.1532 | 0.9354 | 0.9586 | 0.7637 | 20 | 253 | | 1.0967 | 0.5124 | 76 | 1.0131 | 0.0212 | 0.1350 | 0.0402 | 0.2791 | 0.3219 |...

Note8 de ago.huggingface.co/daryaZare/iris-llama-3.2-3b-multihop-only-k10

Source context: 245 downloads · 1 likes · Pipeline text-generation · Library peft · Repo daryaZare/iris-llama-3.2-3b-multihop-only-k10