See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/SmolLM-135M
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - f693d3ab51eecc27_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/f693d3ab51eecc27_train_data.json
  type:
    field_input: span_labels
    field_instruction: masked_text
    field_output: unmasked_text
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 5
eval_max_new_tokens: 128
eval_steps: 200
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: true
hub_model_id: error577/e8bb4bf7-6c39-44cf-980d-b666ffaf3211
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: 64
lora_dropout: 0.15
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
peft_use_rslora: true
lora_target_linear: true
lora_target_modules: 
    - q_proj 
    - v_proj
loraplus_lr_ratio: 16
lr_scheduler: constant_with_warmup
micro_batch_size: 4
mlflow_experiment_name: /tmp/f693d3ab51eecc27_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: ademamix_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
restore_best_weights: true
optim_target_modules: 
    - attn 
    - mlp
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 200
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: 386bd2ee-883e-4683-a1e2-a0d14e23e014
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 386bd2ee-883e-4683-a1e2-a0d14e23e014
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

e8bb4bf7-6c39-44cf-980d-b666ffaf3211

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: 0.9601

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: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADEMAMIX_8BIT and the args are: No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 10
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
2.066	0.0001	1	2.1198
1.4404	0.0154	200	1.0935
1.1616	0.0309	400	1.0196
0.9961	0.0463	600	0.9976
0.9917	0.0618	800	0.9818
1.1452	0.0772	1000	0.9723
1.1945	0.0927	1200	0.9641
0.9992	0.1081	1400	0.9554
1.0653	0.1235	1600	0.9446
1.1229	0.1390	1800	0.9491
1.3471	0.1544	2000	0.9483
1.1403	0.1699	2200	0.9557
1.247	0.1853	2400	0.9642
1.4105	0.2008	2600	0.9601

Framework versions

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