See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/gemma-1.1-2b-it
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 9af2fd9055cb835a_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/9af2fd9055cb835a_train_data.json
  type:
    field_instruction: qwq
    field_output: problem
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 4
eval_max_new_tokens: 128
eval_steps: 150
eval_table_size: null
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/fcc25fb7-e3b9-4bed-a772-9a13d4deab2b
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 128
lora_dropout: 0.3
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 4800
micro_batch_size: 2
mlflow_experiment_name: /tmp/9af2fd9055cb835a_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
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: 150
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.03779232362322565
wandb_entity: null
wandb_mode: online
wandb_name: 22e5cac2-fae3-415f-8da6-ea8f3470cabd
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 22e5cac2-fae3-415f-8da6-ea8f3470cabd
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

fcc25fb7-e3b9-4bed-a772-9a13d4deab2b

This model is a fine-tuned version of unsloth/gemma-1.1-2b-it on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3834

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: 4
• total_train_batch_size: 8
• 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: 4800

Training results

Training Loss	Epoch	Step	Validation Loss
1.7728	0.0001	1	2.0024
0.8611	0.0094	150	0.5627
0.4826	0.0189	300	0.5481
0.7717	0.0283	450	0.5406
1.3571	0.0377	600	0.5391
0.259	0.0471	750	0.5271
0.4037	0.0566	900	0.5219
0.4962	0.0660	1050	0.5143
0.5318	0.0754	1200	0.5031
0.3864	0.0848	1350	0.5021
0.515	0.0943	1500	0.4971
0.7259	0.1037	1650	0.4935
0.4296	0.1131	1800	0.4826
0.3695	0.1225	1950	0.4723
0.2575	0.1320	2100	0.4713
0.3029	0.1414	2250	0.4518
0.6075	0.1508	2400	0.4520
0.4161	0.1602	2550	0.4426
0.3663	0.1697	2700	0.4317
0.3306	0.1791	2850	0.4274
0.5435	0.1885	3000	0.4203
0.5373	0.1980	3150	0.4147
0.3287	0.2074	3300	0.4095
0.4146	0.2168	3450	0.4042
0.3707	0.2262	3600	0.3983
0.4137	0.2357	3750	0.3942
0.4442	0.2451	3900	0.3908
0.4156	0.2545	4050	0.3884
0.3788	0.2639	4200	0.3861
0.4384	0.2734	4350	0.3843
0.2666	0.2828	4500	0.3837
0.2423	0.2922	4650	0.3834
0.3595	0.3016	4800	0.3834

Framework versions

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