Built with Axolotl

See axolotl config

axolotl version: 0.12.2

base_model: Qwen/Qwen3-0.6B
trust_remote_code: true
strict: false

chat_template: qwen3

plugins:
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin

datasets:
  - path: ./Dataset/dataset_bpln.jsonl
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value
    roles:
      user: ["human"]
      assistant: ["gpt"]
      system: ["system"]

dataset_prepared_path: ./process
val_set_size: 0.01
output_dir: ./outputs/out

sequence_len: 256
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true

wandb_project: BPLN
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

adapter: lora
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_modules:
  - q_proj
  - v_proj

load_in_8bit: false
load_in_4bit: false

gradient_accumulation_steps: 4
micro_batch_size: 4
num_epochs: 3

optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 8e-5
weight_decay: 0.0

warmup_ratio: 0.05

bf16: true
fp16: false
tf32: true

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false

flash_attention: false

logging_steps: 1
evals_per_epoch: 1
saves_per_epoch: 1
save_total_limit: 2

special_tokens:
  eos_token: "<|im_end|>"

outputs/out

This model is a fine-tuned version of Qwen/Qwen3-0.6B on the ./Dataset/dataset_bpln.jsonl dataset. It achieves the following results on the evaluation set:

  • Loss: 1.5580
  • Memory/max Mem Active(gib): 1.44
  • Memory/max Mem Allocated(gib): 1.44
  • Memory/device Mem Reserved(gib): 1.49

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: 8e-05
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 16
  • optimizer: Use OptimizerNames.ADAMW_TORCH 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: 4
  • training_steps: 81

Training results

Training Loss Epoch Step Validation Loss Mem Active(gib) Mem Allocated(gib) Mem Reserved(gib)
No log 0 0 1.8057 1.17 1.17 1.19
1.7214 0.9818 27 1.5839 1.44 1.44 1.49
1.7178 1.9455 54 1.5639 1.44 1.44 1.49
1.6562 2.9091 81 1.5580 1.44 1.44 1.49

Framework versions

  • PEFT 0.17.0
  • Transformers 4.55.2
  • Pytorch 2.5.1+cu121
  • Datasets 4.0.0
  • Tokenizers 0.21.4
Downloads last month
-
Safetensors
Model size
0.6B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for LucasMota10/qwen3_0.6B_BPLN

Finetuned
Qwen/Qwen3-0.6B
Adapter
(278)
this model