base_model: NewEden/Hamanasu-KTO-4B
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

hub_model_id: NewEden/Hamanasu-4B-RP
hub_strategy: "all_checkpoints"
push_dataset_to_hub:
hf_use_auth_token: true

plugins:
  - axolotl.integrations.liger.LigerPlugin
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
liger_rope: true
liger_rms_norm: true
liger_layer_norm: true
liger_glu_activation: true
liger_fused_linear_cross_entropy: false
cut_cross_entropy: true

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: NewEden/RP-logs-V2-Experimental-prefixed
    type: dan-chat-advanced
  - path: NewEden/Creative_Writing-Complexity
    type: dan-chat-advanced
  - path: NewEden/Discord-Filtered
    type: dan-chat-advanced
  - path: NewEden/DeepseekRP-Filtered
    type: dan-chat-advanced
  - path: NewEden/Storium-Prefixed-Clean  
    type: dan-chat-advanced
  - path: NewEden/Basket-Weaving-Filtered
    type: dan-chat-advanced
  - path: NewEden/LIMARP-Complexity
    type: dan-chat-advanced
  - path: NewEden/Misc-Data-Sharegpt-Prefixed
    type: dan-chat-advanced
  - path: NewEden/BlueSky-10K-Complexity
    type: dan-chat-advanced 
  - path: NewEden/OpenCAI-ShareGPT
    type: dan-chat-advanced
  - path: NewEden/Basket-Weaving-Filtered
    type: dan-chat-advanced  
  - path: PocketDoc/Dans-Personamaxx-VN
    type: dan-chat-advanced 
  - path: PocketDoc/Dans-Kinomaxx-VanillaBackrooms
    type: dan-chat-advanced  
dataset_prepared_path: prepared_data
val_set_size: 0.0
output_dir: ./output

sequence_len: 32768
sample_packing: true
pad_to_sequence_len: true


wandb_project: tavbussy
wandb_entity:
wandb_watch:
wandb_name: rp-attempt-01
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 2e-5
max_grad_norm: 0.1

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

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

warmup_steps: 40
saves_per_epoch: 2
debug:
deepspeed: deepspeed_configs/zero3_bf16.json
weight_decay: 0.02
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|finetune_right_pad_id|>