base_model: NousResearch/Meta-Llama-3-8B
# optionally might have model_type or tokenizer_type
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer
# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name
pretraining_dataset:
  - path: json
    data_files:
      - data/3.7b.jsonl
  - type: pretrain

# Streaming-specific settings
streaming_multipack_buffer_size: 10000
shuffle_merged_datasets: true

# Training configuration
max_steps: 13000

output_dir: ./outputs/3.7b

sequence_len: 4096
sample_packing: true
# eval_sample_packing: false
pretrain_multipack_attn: true
flash_attention: true


adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_modules_to_save:
  - embed_tokens
  - lm_head

wandb_project: eques-llama
wandb_entity:
wandb_watch:
wandb_name: 0009_3.7b
wandb_log_model:

gradient_accumulation_steps: 6
micro_batch_size: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

bf16: auto
tf32: true

# Logging and checkpointing
logging_steps: 10
save_strategy: steps
save_steps: 500
save_total_limit: 1

warmup_ratio: 0.1
# evals_per_epoch: 4
weight_decay: 0.0
special_tokens:
   pad_token: <|end_of_text|>

# save_first_step: true  # uncomment this to validate checkpoint saving works with your config

outputs/3.7b

This model is a fine-tuned version of NousResearch/Meta-Llama-3-8B on an unknown dataset.

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: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 6
total_train_batch_size: 12
total_eval_batch_size: 2
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: 1300
training_steps: 13000

Training results

Framework versions

PEFT 0.17.0
Transformers 4.55.2
Pytorch 2.6.0+cu126
Datasets 4.0.0
Tokenizers 0.21.4

eques-sec3/3.7b

outputs/3.7b

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

eques-sec3/
3.7b