24b Ms Dans Personality Engine V1.3.0 TestArticle 1

Axolotl版本: 0.10.0.dev0

base_model: Dans-DiscountModels/Mistral-Small-3.1-24B-Base-2503-hf-DanChat
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

trust_remote_code:

# wandb配置
wandb_project: 24b-ms-dans-personality-engine
wandb_watch:

wandb_run_id: V1.3.0-1-5 # V{版本}-{运行编号}-{尝试编号}
wandb_log_model:

# 将检查点推送到Hub
hub_model_id: Dans-DiscountModels/24b-ms-dans-personality-engine-v1.3.0-TestArticle-1
# 如何将检查点推送到Hub
# https://huggingface.co/docs/transformers/v4.31.0/en/main_classes/trainer#transformers.TrainingArguments.hub_strategy
hub_strategy: "every_save"
# 是否使用HF的`use_auth_token`来加载数据集。对于获取私有数据集很有用
# 与`push_dataset_to_hub`结合使用时必须为true
hf_use_auth_token: true

# 完成的模型保存位置
output_dir: ./24b-ms-dans-personality-engine

save_safetensors: true

datasets:
  - path: Dans-DiscountModels/pretokenization-test-6
    ds_type: parquet
    type:

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

adapter:
lora_model_dir:

dataset_prepared_path: ./24b-ms-dans-personality-engine
val_set_size: 0.0

sequence_len: 33000

sample_packing: true
eval_sample_packing: true

pad_to_sequence_len: true

gradient_checkpointing: true

gradient_accumulation_steps: 4
micro_batch_size: 1

num_epochs: 2

optimizer: ademamix_8bit
optim_args: "beta1=0.9,beta2=0.999,beta3=0.999,alpha=5"

lr_scheduler: rex
learning_rate: 0.000001
cosine_min_lr_ratio:

max_grad_norm: 0.001

train_on_inputs: false
group_by_length: false

bf16: true
fp16: false
tf32: false

early_stopping_patience:

resume_from_checkpoint:
auto_resume_from_checkpoints: false

local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_ratio: 0.1

evals_per_epoch: 24
eval_table_size:
eval_max_new_tokens:

saves_per_epoch: 4
save_total_limit: 1

debug: false

deepspeed: deepspeed_configs/zero3_bf16.json

fsdp:
fsdp_config:

special_tokens: