See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_find_batch_size: true
base_model: unsloth/SmolLM2-135M
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - 5b239afff048be33_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/5b239afff048be33_train_data.json
  type:
    field_instruction: instruction
    field_output: output_1
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 1
early_stopping_threshold: 0.001
eval_max_new_tokens: 128
eval_steps: 20
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 2
gradient_checkpointing: false
group_by_length: false
hub_model_id: mrferr3t/ce0c8b53-2ff3-4507-b6e3-fcd3c3e4fd7e
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0005
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 100
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
micro_batch_size: 32
mlflow_experiment_name: /tmp/5b239afff048be33_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 5
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
s2_attention: null
sample_packing: false
save_steps: 20
saves_per_epoch: 0
sequence_len: 512
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 3883aa00-6588-42b3-bf77-b1f07b299789
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 3883aa00-6588-42b3-bf77-b1f07b299789
warmup_ratio: 0.05
weight_decay: 0.0
xformers_attention: null

ce0c8b53-2ff3-4507-b6e3-fcd3c3e4fd7e

This model is a fine-tuned version of unsloth/SmolLM2-135M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5664

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: 0.0005
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use adamw_bnb_8bit 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: 61
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0041	1	1.7893
No log	0.0815	20	1.7588
No log	0.1629	40	1.6881
No log	0.2444	60	1.6510
No log	0.3259	80	1.6331
1.6772	0.4073	100	1.6229
1.6772	0.4888	120	1.6162
1.6772	0.5703	140	1.6114
1.6772	0.6517	160	1.6064
1.6772	0.7332	180	1.6028
1.5962	0.8147	200	1.5993
1.5962	0.8961	220	1.5963
1.5962	0.9776	240	1.5933
1.5962	1.0591	260	1.5926
1.5962	1.1405	280	1.5890
1.5752	1.2220	300	1.5859
1.5752	1.3035	320	1.5832
1.5752	1.3849	340	1.5807
1.5752	1.4664	360	1.5787
1.5752	1.5479	380	1.5769
1.5522	1.6293	400	1.5743
1.5522	1.7108	420	1.5721
1.5522	1.7923	440	1.5704
1.5522	1.8737	460	1.5677
1.5522	1.9552	480	1.5658
1.5454	2.0367	500	1.5664

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.3.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1