See axolotl config

axolotl version: 0.4.1

adapter: qlora
auto_resume_from_checkpoints: false
base_model: unsloth/SmolLM-135M
bf16: auto
chat_template: llama3
dataset_prepared_path: null
dataset_processes: 6
datasets:
- data_files:
  - 0035b2121f3750b5_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/0035b2121f3750b5_train_data.json
  type:
    field_instruction: context
    field_output: question
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 3
eval_max_new_tokens: 128
eval_steps: 200
eval_table_size: null
evals_per_epoch: null
flash_attention: true
fp16: false
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: error577/7a705685-b5a0-46e9-a82b-6b03715fff7d
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 64
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: null
micro_batch_size: 10
mlflow_experiment_name: /tmp/0035b2121f3750b5_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_torch_4bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 200
sequence_len: 512
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: 1661de9e-91d5-48bc-825d-83583560bcf1
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 1661de9e-91d5-48bc-825d-83583560bcf1
warmup_steps: 30
weight_decay: 0.0
xformers_attention: null

7a705685-b5a0-46e9-a82b-6b03715fff7d

This model is a fine-tuned version of unsloth/SmolLM-135M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.2724

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: 10
• eval_batch_size: 10
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 40
• optimizer: Use OptimizerNames.ADAMW_TORCH_4BIT 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: 30
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
4.0837	0.0004	1	3.9034
2.0119	0.0888	200	2.0735
1.9729	0.1775	400	1.8873
1.7298	0.2663	600	1.7849
1.6232	0.3551	800	1.7007
1.8815	0.4439	1000	1.6521
1.7189	0.5326	1200	1.6111
1.6542	0.6214	1400	1.5725
1.5466	0.7102	1600	1.5364
1.5247	0.7989	1800	1.5042
1.4437	0.8877	2000	1.4765
1.3985	0.9765	2200	1.4447
1.3636	1.0652	2400	1.4302
1.4208	1.1540	2600	1.4131
1.3557	1.2428	2800	1.3957
1.4026	1.3316	3000	1.3781
1.1675	1.4203	3200	1.3662
1.2471	1.5091	3400	1.3605
1.3073	1.5979	3600	1.3409
1.2123	1.6866	3800	1.3271
1.342	1.7754	4000	1.3176
1.2351	1.8642	4200	1.3085
1.2137	1.9530	4400	1.3016
1.3767	2.0417	4600	1.3004
1.1753	2.1305	4800	1.2907
1.2174	2.2193	5000	1.2861
1.2507	2.3080	5200	1.2850
1.2156	2.3968	5400	1.2805
1.2422	2.4856	5600	1.2778
1.3742	2.5743	5800	1.2753
1.2677	2.6631	6000	1.2738
1.3482	2.7519	6200	1.2720
1.1884	2.8407	6400	1.2724
1.3061	2.9294	6600	1.2724

Framework versions

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