base_model: Qwen/Qwen2.5-32B
datasets:
  - path: StudyPal/education
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value
output_dir: ./outputs/qwen25-croatian-education

sequence_len: 1024
adapter: lora

lora_r: 32
lora_alpha: 64
lora_dropout: 0.05
lora_target_modules:
  - q_proj
  - v_proj
  - k_proj
  - o_proj
  - gate_proj
  - down_proj
  - up_proj

gradient_accumulation_steps: 8
micro_batch_size: 2
max_steps: 1500
optimizer: paged_adamw_8bit
learning_rate: 0.0001
load_in_4bit: true
train_on_inputs: false
bf16: auto

# Additional settings for 32B model
trust_remote_code: true
bnb_4bit_quant_type: nf4
bnb_4bit_use_double_quant: true
bnb_4bit_compute_dtype: bfloat16
gradient_checkpointing: true
sample_packing: true
pad_to_sequence_len: true
val_set_size: 0.001
warmup_steps: 75
save_safetensors: true
flash_attention: true
dataloader_num_workers: 8
dataloader_pin_memory: true
eval_steps: 999999
eval_strategy: steps
save_steps: 750
logging_steps: 50

special_tokens:
  bos_token: "<|im_start|>"
  eos_token: "<|im_end|>"
  pad_token: "<|endoftext|>"

outputs/qwen25-croatian-education

This model is a fine-tuned version of Qwen/Qwen2.5-32B on the StudyPal/education dataset.

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: 0.0001
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 16
optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT 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: 75
training_steps: 1500

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0003	1	1.4631

Framework versions

PEFT 0.14.0
Transformers 4.49.0
Pytorch 2.5.1+cu124
Datasets 3.2.0
Tokenizers 0.21.0

aerodynamics21/StudyPal-LLM-1

outputs/qwen25-croatian-education

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

aerodynamics21/
StudyPal-LLM-1