adapter: lora
base_model: NousResearch/Yarn-Mistral-7b-128k
bf16: true
chat_template: llama3
datasets:
- data_files:
  - 12015d7c9ee7f3df_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/
  type:
    field_input: None
    field_instruction: instruct
    field_output: output
    field_system: None
    format: None
    no_input_format: None
    system_format: '{system}'
    system_prompt: None
eval_max_new_tokens: 256
evals_per_epoch: 2
flash_attention: false
fp16: false
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: true
hub_model_id: apriasmoro/27e554a7-9349-41b8-b91f-45cc2482a433
learning_rate: 0.0002
logging_steps: 10
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: false
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
max_steps: 15
micro_batch_size: 12
mlflow_experiment_name: /tmp/12015d7c9ee7f3df_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
sample_packing: false
save_steps: 200
sequence_len: 2048
special_tokens:
  pad_token: </s>
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 0aa91fdd-f464-4c35-9e87-5ba2524c6ecc
wandb_project: Gradients-On-Demand
wandb_run: apriasmoro
wandb_runid: 0aa91fdd-f464-4c35-9e87-5ba2524c6ecc
warmup_steps: 100
weight_decay: 0.01

27e554a7-9349-41b8-b91f-45cc2482a433

This model is a fine-tuned version of NousResearch/Yarn-Mistral-7b-128k on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 1.5053

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: 12
eval_batch_size: 12
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 48
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: 100
training_steps: 15

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0702	1	1.5261
No log	0.2105	3	1.5915
No log	0.4211	6	1.5176
No log	0.6316	9	1.4834
2.1415	0.8421	12	1.4475
2.1415	1.0	15	1.5053

Framework versions

PEFT 0.15.2
Transformers 4.51.3
Pytorch 2.5.1+cu124
Datasets 3.5.1
Tokenizers 0.21.1