See axolotl config

axolotl version: 0.4.1

base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
# I'm training on 4090 GPUs
# so I'm using 4-bit precision to save on memory
load_in_4bit: true
strict: false

data_seed: 42
seed: 42

datasets:
  - path: data/sharegpt_isaf_press_releases_ft_train.jsonl
    type: sharegpt
    conversation: alpaca
dataset_prepared_path:
val_set_size: 0.1
output_dir: ./outputs/tiny-llama/lora-out-sharegpt
hub_model_id: strickvl/isafpr-tiny-llama-lora-sharegpt

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

adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

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

gradient_accumulation_steps: 4
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_steps: 10
evals_per_epoch: 4
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"

isafpr-tiny-llama-lora-sharegpt

This model is a fine-tuned version of TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0507

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: 4
• total_train_batch_size: 16
• 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: 10
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss
1.7687	0.0270	1	1.7719
1.0632	0.2703	10	0.9033
0.1374	0.5405	20	0.1365
0.0763	0.8108	30	0.0942
0.0752	1.0608	40	0.0765
0.0764	1.3311	50	0.0680
0.0623	1.6014	60	0.0630
0.0596	1.8716	70	0.0593
0.0523	2.1216	80	0.0570
0.0514	2.3919	90	0.0543
0.0501	2.6622	100	0.0528
0.0475	2.9324	110	0.0515
0.0525	3.1824	120	0.0511
0.0436	3.4527	130	0.0509
0.0508	3.7230	140	0.0507

Framework versions

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