YoungPanda / SFT_Test1

axolotl version: 0.4.0

base_model: 152334H/miqu-1-70b-sf
model_type: AutoModelForCausalLM
tokenizer_type: LlamaTokenizer
trust_remote_code: true

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: Drewskidang/rag
    type: alpaca
    prompt_style: chatml
  - path: Drewskidang/output
    type: alpaca
    prompt_style: chatml
  - path: Drewskidang/Instruct
    type: alpaca
    prompt_style: chatml
  - path: Drewskidang/Preference
    type: alpaca
    prompt_style: chatml
  - path: lighteval/legal_summarization
    name: BillSum
    type: summarizetldr

  
dataset_prepared_path: last_run_prepared
val_set_size: 0.0
output_dir: ./qlora-out

## You can optionally freeze the entire model and unfreeze a subset of parameters
unfrozen_parameters:
  - lm_head.*
  #- model.embed_tokens.*
#- model.layers.2[0-9]+.block_sparse_moe.gate.*
#  - model.layers.2[0-9]+.block_sparse_moe.experts.*
#  - model.layers.3[0-9]+.block_sparse_moe.gate.*
#  - model.layers.3[0-9]+.block_sparse_moe.experts.*

model_config:
  output_router_logits: true



sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

wandb_project: Leak_Mistral
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 10
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

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: 10
evals_per_epoch: 4
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed: 
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  eos_token: "<|im_end|>"
tokens:
  - "<|im_start|>"
trust_remote_code: true