See axolotl config

axolotl version: 0.4.1

adapter: qlora
base_model: unsloth/SmolLM-135M-Instruct
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - ebcebb14d0612ac2_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/ebcebb14d0612ac2_train_data.json
  type:
    field_instruction: link
    field_output: text
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 1
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: error577/78e02d3e-8f98-4c6a-b5d5-cf8895e1e778
hub_repo: null
hub_strategy: end
hub_token: null
learning_rate: 0.0001
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
max_steps: 500
micro_batch_size: 1
mlflow_experiment_name: /tmp/ebcebb14d0612ac2_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 20
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
saves_per_epoch: 1
sequence_len: 1024
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.002
wandb_entity: null
wandb_mode: online
wandb_name: f32f4f56-4903-4188-a885-31957e402d25
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: f32f4f56-4903-4188-a885-31957e402d25
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

78e02d3e-8f98-4c6a-b5d5-cf8895e1e778

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

• Loss: 2.2006

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• 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: 500

Training results

Training Loss	Epoch	Step	Validation Loss
2.9183	0.0000	1	2.7664
2.8161	0.0004	25	2.6765
2.5902	0.0008	50	2.5414
2.467	0.0012	75	2.4589
2.4378	0.0016	100	2.3975
2.2701	0.0020	125	2.3515
2.4121	0.0023	150	2.3166
2.2649	0.0027	175	2.2896
2.3924	0.0031	200	2.2701
2.3039	0.0035	225	2.2538
2.3156	0.0039	250	2.2400
2.3732	0.0043	275	2.2300
2.3055	0.0047	300	2.2209
2.2445	0.0051	325	2.2144
2.5038	0.0055	350	2.2094
2.2085	0.0059	375	2.2060
2.0684	0.0063	400	2.2032
2.085	0.0066	425	2.2017
2.3817	0.0070	450	2.2009
2.1274	0.0074	475	2.2008
2.3271	0.0078	500	2.2006

Framework versions

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