CAI Supernova R2

Axolotl 版本：0.4.1

base_model: arcee-ai/Llama-3.1-SuperNova-Lite
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: NewEden/CharacterAI-logs-sharegpt-Ngram-Cleaned
    type: sharegpt
    conversation: llama3
  - path: NewEden/OpenCAI-ShareGPT
    type: sharegpt
    conversation: llama3

chat_template: llama3

  #val_set_size: 0.01
output_dir: ./outputs

adapter:
lora_r:
lora_alpha:
lora_dropout:
lora_target_linear:

sequence_len: 16384
# sequence_len: 32768
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true

wandb_project: CAI-Supernova
wandb_entity:
wandb_watch:
wandb_name: CAI-Supernova-2
wandb_log_model:

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_swiglu: true
liger_fused_linear_cross_entropy: true

gradient_accumulation_steps: 2
micro_batch_size: 1
num_epochs: 4
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 1e-5
weight_decay: 0.05

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

gradient_checkpointing: unsloth
early_stopping_patience:
resume_from_checkpoint:
#auto_resume_from_checkpoints: true
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 15
  #evals_per_epoch: 4
eval_table_size:
  #eval_max_new_tokens: 128
saves_per_epoch: 1

debug:
deepspeed: /workspace/axolotl/deepspeed_configs/zero3_bf16_cpuoffload_params.json
fsdp:
fsdp_config:

special_tokens:
  pad_token: <|finetune_right_pad_id|>
  eos_token: <|eot_id|>