poti-wp-llama3-8b-lora

<!-- 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. -->

<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>

<details><summary>See axolotl config</summary>

axolotl version: 0.13.0.dev0

# ============================================
# 🔥 POTI-WP - MODO ESTÁVEL MAS FORTE 🔥
# ============================================

base_model: unsloth/llama-3-8b-bnb-4bit
model_type: LlamaForCausalLM
load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: mrmsoares/poti-wp-instruct
    type: alpaca
    shards: 1

dataset_prepared_path: last_run_prepared
val_set_size: 0.02
output_dir: ./outputs/poti-wp-lora-chernobyl

adapter: qlora

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: false

lora_r: 64
lora_alpha: 128
lora_dropout: 0.05
lora_target_linear: true

# Desabilita os kernels otimizados que estão dando problema
lora_mlp_kernel: false
lora_qkv_kernel: false
lora_o_kernel: false

micro_batch_size: 2
gradient_accumulation_steps: 8

num_epochs: 4
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 2e-4
train_on_inputs: false
group_by_length: false

bf16: auto
fp16: false
tf32: true

gradient_checkpointing: true

flash_attention: true

torch_compile: false

dataloader_num_workers: 2
dataloader_pin_memory: true

logging_steps: 1
warmup_steps: 10
evals_per_epoch: 2
saves_per_epoch: 1

weight_decay: 0.01
seed: 42
neftune_noise_alpha: 5

</details><br>

outputs/poti-wp-lora-chernobyl

This model is a fine-tuned version of unsloth/llama-3-8b-bnb-4bit on the mrmsoares/poti-wp-instruct dataset. It achieves the following results on the evaluation set:

• Loss: 12.6679
• Ppl: 317384.6004
• Memory/max Active (gib): 12.45
• Memory/max Allocated (gib): 12.45
• Memory/device Reserved (gib): 16.26

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: 0.0002
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Use 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: 10
• training_steps: 16

Training results

Framework versions

• PEFT 0.18.1.dev0
• Transformers 4.57.1
• Pytorch 2.8.0+cu128
• Datasets 4.4.2
• Tokenizers 0.22.1