base_model: ./meta-llama_Llama-3.1-8B
# 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

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: 2025-01_conversations_2024_2.jsonl
    type: chat_template
    chat_template: tokenizer_default
    field_messages: conversations
    message_field_role: from
    message_field_content: value
    roles:
      <|autheur|>:
        - human
      <|khey|>:
        - gpt
      <|sujet|>:
        - system
dataset_prepared_path:
val_set_size: 0.05
output_dir: ./outputs/
dataset_prepared_path: last_run_prepared

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

wandb_project: JVCGPT Medium 8b v2
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 2
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 1e-5

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

gradient_checkpointing: unsloth
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
s2_attention:

warmup_steps: 100
eval_table_size:
saves_per_epoch: 20
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
   pad_token: <|end_of_text|>
save_safetensors: true
save_total_limit: 5
resume_from_checkpoint: ./outputs/checkpoint-9999
# If resume_from_checkpoint isn't set and you simply want it to start where it left off.
# Be careful with this being turned on between different models.
auto_resume_from_checkpoints: true