See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2.5-3B
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 8e23ab7f3136aade_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/8e23ab7f3136aade_train_data.json
  type:
    field_instruction: input persona
    field_output: synthesized text
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 5
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: "unsloth"
group_by_length: true
hub_model_id: error577/f44fd2c7-eff5-4055-a260-701e9f4bccfb
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_in_4bit: false
load_in_8bit: true
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.15
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
loraplus_lr_ratio: 8
lr_scheduler: cosine
micro_batch_size: 1
mlflow_experiment_name: /tmp/8e23ab7f3136aade_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: paged_ademamix_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
restore_best_weights: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 100
sequence_len: 1024
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.005
wandb_entity: null
wandb_mode: online
wandb_name: 996763c0-8994-4543-842c-7a5c088c235f
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 996763c0-8994-4543-842c-7a5c088c235f
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

f44fd2c7-eff5-4055-a260-701e9f4bccfb

This model is a fine-tuned version of unsloth/Qwen2.5-3B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8224

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: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.PAGED_ADEMAMIX_8BIT and the args are: No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
0.5376	0.0002	1	1.1038
1.0376	0.0164	100	0.8196
0.8848	0.0328	200	0.8143
0.9346	0.0492	300	0.8134
0.8966	0.0656	400	0.8115
1.008	0.0820	500	0.8160
0.8225	0.0985	600	0.8154
0.857	0.1149	700	0.8197
1.0713	0.1313	800	0.8197
1.0041	0.1477	900	0.8224

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1