See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2.5-1.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - a88e78e41748bf83_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/a88e78e41748bf83_train_data.json
  type:
    field_instruction: prompt
    field_output: generation
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
device_map:
  ? ''
  : 0,1,2,3,4,5,6,7
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/6baccfbf-bbbc-4bc2-a86c-e8d2a31c938c
hub_repo: null
hub_strategy: null
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 2007
micro_batch_size: 4
mlflow_experiment_name: /tmp/a88e78e41748bf83_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 2
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 100
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: 895387c3-0be7-49d8-a314-abeba9f636b4
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 895387c3-0be7-49d8-a314-abeba9f636b4
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

6baccfbf-bbbc-4bc2-a86c-e8d2a31c938c

This model is a fine-tuned version of unsloth/Qwen2.5-1.5B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7610

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_BNB 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: 2007

Training results

Training Loss	Epoch	Step	Validation Loss
1.7715	0.0010	1	1.7711
0.9367	0.0963	100	0.9125
0.9026	0.1926	200	0.8665
0.865	0.2888	300	0.8448
0.8792	0.3851	400	0.8292
0.8294	0.4814	500	0.8173
0.8027	0.5777	600	0.8093
0.8164	0.6740	700	0.8003
0.8412	0.7702	800	0.7946
0.7625	0.8665	900	0.7882
0.7594	0.9628	1000	0.7830
0.7531	1.0596	1100	0.7792
0.7919	1.1559	1200	0.7752
0.71	1.2521	1300	0.7721
0.7058	1.3484	1400	0.7684
0.7522	1.4447	1500	0.7661
0.7818	1.5410	1600	0.7639
0.7388	1.6373	1700	0.7625
0.6826	1.7335	1800	0.7616
0.7884	1.8298	1900	0.7611
0.777	1.9261	2000	0.7610

Framework versions

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