Hugging Face's logo

Fischerboot
/
llama3-sophie-adapter-model

PEFT
llama
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Community
llama3-sophie-adapter-model / README.md
Fischerboot's picture
Fischerboot
uploaded ze files
368c15b verified
preview code
|
raw
history blame contribute delete
13.8 kB
metadata
base_model: Fischerboot/LLama3-Lexi-Aura-3Some-SLERP-SLERP-ql-merge
library_name: peft
tags:
  - generated_from_trainer
model-index:
  - name: outputs/qlora-out
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.4.1

base_model: Fischerboot/LLama3-Lexi-Aura-3Some-SLERP-SLERP-ql-merge
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

chat_template: llama3
datasets:
  - path: Fischerboot/dataset
    type: sharegpt
    conversation: llama3
dataset_prepared_path: last_run_prepared
val_set_size: 0.1
output_dir: ./outputs/qlora-out

adapter: qlora
lora_model_dir:

sequence_len: 128
sample_packing: false
pad_to_sequence_len: true

lora_r: 1024
lora_alpha: 512
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 50
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

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

eval_sample_packing: false
warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<|begin_of_text|>"
  eos_token: "<|end_of_text|>"
  pad_token: "<|end_of_text|>"

outputs/qlora-out

This model is a fine-tuned version of Fischerboot/LLama3-Lexi-Aura-3Some-SLERP-SLERP-ql-merge on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0000

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: 0.0002
  • train_batch_size: 1
  • eval_batch_size: 1
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 50

Training results

Training Loss Epoch Step Validation Loss
5.9056 0.0030 1 5.7722
0.0002 0.25 82 0.0002
0.0262 0.5 164 0.0310
0.0004 0.75 246 0.1491
0.2739 1.0 328 0.0013
0.0005 1.25 410 0.0490
0.0 1.5 492 0.0000
0.0 1.75 574 0.0000
0.0 2.0 656 0.0000
0.0 2.25 738 0.0000
0.0 2.5 820 0.0000
0.0 2.75 902 0.0000
0.0 3.0 984 0.0000
0.0 3.25 1066 0.0000
0.0 3.5 1148 0.0000
0.0 3.75 1230 0.0000
0.0 4.0 1312 0.0000
0.0 4.25 1394 0.0000
0.0 4.5 1476 0.0000
0.0 4.75 1558 0.0000
0.0 5.0 1640 0.0000
0.0 5.25 1722 0.0000
0.0 5.5 1804 0.0000
0.0 5.75 1886 0.0000
0.0 6.0 1968 0.0000
0.0 6.25 2050 0.0000
0.0 6.5 2132 0.0000
0.0 6.75 2214 0.0000
0.0 7.0 2296 0.0000
0.0 7.25 2378 0.0000
0.0 7.5 2460 0.0000
0.0 7.75 2542 0.0000
0.0 8.0 2624 0.0000
0.0 8.25 2706 0.0000
0.0 8.5 2788 0.0000
0.0 8.75 2870 0.0000
0.0 9.0 2952 0.0000
0.0 9.25 3034 0.0000
0.0 9.5 3116 0.0000
0.0 9.75 3198 0.0000
0.0 10.0 3280 0.0000
0.0 10.25 3362 0.0000
0.0 10.5 3444 0.0000
0.0 10.75 3526 0.0000
0.0 11.0 3608 0.0000
0.0 11.25 3690 0.0000
0.0 11.5 3772 0.0000
0.0 11.75 3854 0.0000
0.0 12.0 3936 0.0000
0.0 12.25 4018 0.0000
0.0 12.5 4100 0.0000
0.0 12.75 4182 0.0000
0.0 13.0 4264 0.0000
0.0 13.25 4346 0.0000
0.0 13.5 4428 0.0000
0.0 13.75 4510 0.0000
0.0 14.0 4592 0.0000
0.0 14.25 4674 0.0000
0.0 14.5 4756 0.0000
0.0 14.75 4838 0.0000
0.0 15.0 4920 0.0000
0.0 15.25 5002 0.0000
0.0 15.5 5084 0.0000
0.0 15.75 5166 0.0000
0.0 16.0 5248 0.0000
0.0 16.25 5330 0.0000
0.0 16.5 5412 0.0000
0.0 16.75 5494 0.0000
0.0 17.0 5576 0.0000
0.0 17.25 5658 0.0000
0.0 17.5 5740 0.0000
0.0 17.75 5822 0.0000
0.0 18.0 5904 0.0000
0.0 18.25 5986 0.0000
0.0 18.5 6068 0.0000
0.0 18.75 6150 0.0000
0.0 19.0 6232 0.0000
0.0 19.25 6314 0.0000
0.0 19.5 6396 0.0000
0.0 19.75 6478 0.0000
0.0 20.0 6560 0.0000
0.0 20.25 6642 0.0000
0.0 20.5 6724 0.0000
0.0 20.75 6806 0.0000
0.0 21.0 6888 0.0000
0.0 21.25 6970 0.0000
0.0 21.5 7052 0.0000
0.0 21.75 7134 0.0000
0.0 22.0 7216 0.0000
0.0 22.25 7298 0.0000
0.0 22.5 7380 0.0000
0.0 22.75 7462 0.0000
0.0 23.0 7544 0.0000
0.0 23.25 7626 0.0000
0.0 23.5 7708 0.0000
0.0 23.75 7790 0.0000
0.0 24.0 7872 0.0000
0.0 24.25 7954 0.0000
0.0 24.5 8036 0.0000
0.0 24.75 8118 0.0000
0.0 25.0 8200 0.0000
0.0 25.25 8282 0.0000
0.0 25.5 8364 0.0000
0.0 25.75 8446 0.0000
0.0 26.0 8528 0.0000
0.0 26.25 8610 0.0000
0.0 26.5 8692 0.0000
0.0 26.75 8774 0.0000
0.0 27.0 8856 0.0000
0.0 27.25 8938 0.0000
0.0 27.5 9020 0.0000
0.0 27.75 9102 0.0000
0.0 28.0 9184 0.0000
0.0 28.25 9266 0.0000
0.0 28.5 9348 0.0000
0.0 28.75 9430 0.0000
0.0 29.0 9512 0.0000
0.0 29.25 9594 0.0000
0.0 29.5 9676 0.0000
0.0 29.75 9758 0.0000
0.0 30.0 9840 0.0000
0.0 30.25 9922 0.0000
0.0 30.5 10004 0.0000
0.0 30.75 10086 0.0000
0.0 31.0 10168 0.0000
0.0 31.25 10250 0.0000
0.0 31.5 10332 0.0000
0.0 31.75 10414 0.0000
0.0 32.0 10496 0.0000
0.0 32.25 10578 0.0000
0.0 32.5 10660 0.0000
0.0 32.75 10742 0.0000
0.0 33.0 10824 0.0000
0.0 33.25 10906 0.0000
0.0 33.5 10988 0.0000
0.0 33.75 11070 0.0000
0.0 34.0 11152 0.0000
0.0 34.25 11234 0.0000
0.0 34.5 11316 0.0000
0.0 34.75 11398 0.0000
0.0 35.0 11480 0.0000
0.0 35.25 11562 0.0000
0.0 35.5 11644 0.0000
0.0 35.75 11726 0.0000
0.0 36.0 11808 0.0000
0.0 36.25 11890 0.0000
0.0 36.5 11972 0.0000
0.0 36.75 12054 0.0000
0.0 37.0 12136 0.0000
0.0 37.25 12218 0.0000
0.0 37.5 12300 0.0000
0.0 37.75 12382 0.0000
0.0 38.0 12464 0.0000
0.0 38.25 12546 0.0000
0.0 38.5 12628 0.0000
0.0 38.75 12710 0.0000
0.0 39.0 12792 0.0000
0.0 39.25 12874 0.0000
0.0 39.5 12956 0.0000
0.0 39.75 13038 0.0000
0.0 40.0 13120 0.0000
0.0 40.25 13202 0.0000
0.0 40.5 13284 0.0000
0.0 40.75 13366 0.0000
0.0 41.0 13448 0.0000
0.0 41.25 13530 0.0000
0.0 41.5 13612 0.0000
0.0 41.75 13694 0.0000
0.0 42.0 13776 0.0000
0.0 42.25 13858 0.0000
0.0 42.5 13940 0.0000
0.0 42.75 14022 0.0000
0.0 43.0 14104 0.0000
0.0 43.25 14186 0.0000
0.0 43.5 14268 0.0000
0.0 43.75 14350 0.0000
0.0 44.0 14432 0.0000
0.0 44.25 14514 0.0000
0.0 44.5 14596 0.0000
0.0 44.75 14678 0.0000
0.0 45.0 14760 0.0000
0.0 45.25 14842 0.0000
0.0 45.5 14924 0.0000
0.0 45.75 15006 0.0000
0.0 46.0 15088 0.0000
0.0 46.25 15170 0.0000
0.0 46.5 15252 0.0000
0.0 46.75 15334 0.0000
0.0 47.0 15416 0.0000
0.0 47.25 15498 0.0000
0.0 47.5 15580 0.0000
0.0 47.75 15662 0.0000
0.0 48.0 15744 0.0000
0.0 48.25 15826 0.0000
0.0 48.5 15908 0.0000
0.0 48.75 15990 0.0000
0.0 49.0 16072 0.0000
0.0 49.25 16154 0.0000
0.0 49.5 16236 0.0000
0.0 49.75 16318 0.0000
0.0 50.0 16400 0.0000

Framework versions

  • PEFT 0.11.1
  • Transformers 4.41.1
  • Pytorch 2.1.2+cu118
  • Datasets 2.19.1
  • Tokenizers 0.19.1