File size: 1,419 Bytes
d0579e3 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 |
base_model: huggyllama/llama-13b
base_model_config: huggyllama/llama-13b
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer
load_in_8bit: false
datasets:
- path: anon8231489123/ShareGPT_Vicuna_unfiltered
data_files: ShareGPT_V3_unfiltered_cleaned_split_no_imsorry.json
type: sharegpt
- path: ehartford/WizardLM_alpaca_evol_instruct_70k_unfiltered
type: alpaca
- path: ehartford/wizard_vicuna_70k_unfiltered
type: sharegpt
dataset_prepared_path: last_run_prepared
val_set_size: 0.02
adapter:
lora_model_dir:
sequence_len: 2048
max_packed_sequence_len: 2048
lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
lora_target_modules:
- q_proj
- v_proj
lora_fan_in_fan_out: false
wandb_project: wizard-mega-13b
wandb_watch:
wandb_run_id:
wandb_log_model:
output_dir: ./wizard-mega-13b
batch_size: 512
micro_batch_size: 8
num_epochs: 3
optimizer:
torchdistx_path:
lr_scheduler:
learning_rate: 0.00006
train_on_inputs: false
group_by_length: false
bf16: true
tf32: true
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention: true
flash_attention:
gptq_groupsize:
gptq_model_v1:
warmup_steps: 20
eval_steps: 10
save_steps:
debug:
deepspeed:
weight_decay: 0
fsdp:
- full_shard
- auto_wrap
fsdp_config:
fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
tokens:
bos_token: "<s>"
eos_token: "</s>"
unk_token: "<unk>"
|