base_model: Unbabel/TowerInstruct-7B-v0.2
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer
is_llama_derived_model: true

load_in_8bit: false
load_in_4bit: false
strict: false

device_map: null

datasets:
  - path: /ikerlariak/igarcia945/Mortadelo-Filemon/translation_data/translation_instruction_axolotl.jsonl
    type: sharegpt
    conversation: chatml
    field: conversations
dataset_prepared_path:
val_set_size: 0.05

output_dir: /ikerlariak/igarcia945/Mortadelo-Filemon/Tower7B-EN2ES

adapter: lora
lora_model_dir:

sequence_len: 8096
sample_packing: false
eval_sample_packing: false
pad_to_sequence_len: false

overrides_of_model_config:
  # RoPE Scaling https://github.com/huggingface/transformers/pull/24653
  rope_scaling:
    type: dynamic
    factor: 2.0


lora_r: 128
lora_alpha: 256
lora_dropout: 0.05
lora_target_modules: 
lora_target_linear: true
lora_fan_in_fan_out:
lora_modules_to_save:
  - embed_tokens
  - lm_head

wandb_project: Mortadelo&Filemon
wandb_entity: igarciaf
wandb_watch:
wandb_name: Tower7B-EN2ES
wandb_log_model: 

gradient_accumulation_steps: 8
micro_batch_size: 2
eval_batch_size: 2
num_epochs: 3
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_ratio: 0.03
evals_per_epoch: 2
eval_table_size:
save_strategy: "no"
debug:
deepspeed: /ikerlariak/igarcia945/Mortadelo-Filemon/train_configs/deepspeed_zero3.json
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens: