See axolotl config

axolotl version: 0.4.0

base_model: augmxnt/shisa-base-7b-v1
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

hub_model_id: yentinglin/shisa-7b-v1-sharegpt
hub_strategy: end

datasets:
  - path: NTQAI/sharegpt-clean-ja
    type: sharegpt
    conversation: chatml

chat_template: chatml
dataset_prepared_path: last_run_prepared
val_set_size: 0
output_dir: ./output/ja/sft/shisa-7b-v1/sharegpt/

sequence_len: 4096
sample_packing: false
eval_sample_packing: false
pad_to_sequence_len: true

adapter:
lora_model_dir:
lora_r:
lora_alpha:
lora_dropout:
lora_target_linear:
lora_fan_in_fan_out:

wandb_project: JA-LLM
wandb_entity:
wandb_watch:
wandb_name: sft-fft-sharegpt-clean-ja
wandb_run_id:
wandb_log_model:

gradient_accumulation_steps: 8
micro_batch_size: 4
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
cosine_min_lr_ratio: 0.1 # decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of peak lr
learning_rate: 1e-5
adam_beta1: 0.9
adam_beta2: 0.95
adam_eps: 0.00001
max_grad_norm: 1.0

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: 5
xformers_attention:
flash_attention: true
flash_attn_cross_entropy: false
flash_attn_rms_norm: true
flash_attn_fuse_qkv: false
flash_attn_fuse_mlp: true

warmup_ratio: 0.02  # cannot use with warmup_steps
evals_per_epoch: 1
eval_table_size:
save_per_epoch: 1
save_total_limit: 1
debug:
deepspeed: deepspeed_configs/zero1.json # multi-gpu only
weight_decay: 0.001
fsdp:
fsdp_config:
special_tokens:

ddp_timeout: 180000

special_tokens:
  eos_token: "<|im_end|>"
tokens:
  - "<|im_start|>"

shisa-7b-v1-sharegpt

This model is a fine-tuned version of augmxnt/shisa-base-7b-v1 on the None dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 256
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 4

Training results

Framework versions

• Transformers 4.40.0.dev0
• Pytorch 2.1.2+cu118
• Datasets 2.18.0
• Tokenizers 0.15.0