See axolotl config

axolotl version: 0.4.1

# use google/gemma-7b if you have access
base_model: google/gemma-2b
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: data/templatefree_isaf_press_releases_ft_train.jsonl
    type: input_output
val_set_size: 0.1
output_dir: ./outputs/gemma/qlora-out-templatefree
hub_model_id: strickvl/isafpr-gemma-qlora-templatefree

adapter: qlora
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_modules_to_save:
 - embed_tokens
 - lm_head

sequence_len: 1024
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true

wandb_project: isaf_pr_ft
wandb_entity: strickvl
wandb_watch:
wandb_name:
wandb_log_model:


gradient_accumulation_steps: 3
micro_batch_size: 2
num_epochs: 4
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_ratio: 0.1
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: "<s>"
  eos_token: "</s>"

isafpr-gemma-qlora-templatefree

This model is a fine-tuned version of google/gemma-2b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0379

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: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 3
• total_train_batch_size: 12
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 64
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss
2.3995	0.0054	1	2.3804
0.1051	0.2527	47	0.0906
0.0444	0.5054	94	0.0617
0.0292	0.7581	141	0.0490
0.1049	1.0108	188	0.0475
0.03	1.2419	235	0.0435
0.0219	1.4946	282	0.0411
0.0286	1.7473	329	0.0413
0.0403	2.0	376	0.0383
0.0274	2.2330	423	0.0386
0.0178	2.4857	470	0.0384
0.0272	2.7384	517	0.0378
0.0409	2.9910	564	0.0371
0.013	3.2240	611	0.0378
0.0177	3.4767	658	0.0380
0.018	3.7294	705	0.0379

Framework versions

• PEFT 0.11.1
• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1