See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: echarlaix/tiny-random-mistral
bf16: auto
chat_template: chatml
dataset_prepared_path: null
datasets:
- data_files:
  - a74ecd5c5b3909f6_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/a74ecd5c5b3909f6_train_data.json
  type:
    field_instruction: prompt
    field_output: reference_response
    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: 4
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: false
group_by_length: false
hub_model_id: ardaspear/35064bc1-2c15-4036-bbb1-561a74589740
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: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_steps: 50
micro_batch_size: 2
mlflow_experiment_name: /tmp/a74ecd5c5b3909f6_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
saves_per_epoch: 4
sequence_len: 4056
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: leixa-personal
wandb_mode: online
wandb_name: 35064bc1-2c15-4036-bbb1-561a74589740
wandb_project: Gradients-On-Two
wandb_run: your_name
wandb_runid: 35064bc1-2c15-4036-bbb1-561a74589740
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

35064bc1-2c15-4036-bbb1-561a74589740

This model is a fine-tuned version of echarlaix/tiny-random-mistral on the None dataset. It achieves the following results on the evaluation set:

• Loss: 10.3595

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: 50

Training results

Training Loss	Epoch	Step	Validation Loss
41.5398	0.0002	1	10.3783
41.5426	0.0008	5	10.3779
41.5231	0.0016	10	10.3762
41.4979	0.0024	15	10.3736
41.4805	0.0033	20	10.3706
41.4671	0.0041	25	10.3673
41.4543	0.0049	30	10.3643
41.4492	0.0057	35	10.3618
41.4506	0.0065	40	10.3603
41.4457	0.0073	45	10.3597
41.4237	0.0082	50	10.3595

Framework versions

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