Lambent/CosmoAlpacaLisa-0.2-1b

Trying out some LISA training. This one used the same learning rate as the LORA training and only 4 layers each 10 steps. Honestly these numbers are probably noise, with how close they are.

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
CosmoAlpacaLisa-0.2-1b	23.81	51.75	39.31	29.04	35.98

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
CosmoAlpacaLisa-1b	23.89	51.93	39.93	28.68	36.11

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
CosmoAlpacaLight-1b	24.28	51.31	40.33	29.47	36.35

Model	AGIEval	GPT4All	TruthfulQA	Bigbench	Average
cosmo-1b	22.97	52.01	38.02	28.73	35.43

See axolotl config

axolotl version: 0.4.0

base_model: HuggingFaceTB/cosmo-1b
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: vicgalle/alpaca-gpt4
    type: alpaca
dataset_prepared_path:
val_set_size: 0.05
output_dir: ./lisa-out

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true

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

lisa_n_layers: 4
lisa_step_interval: 10
lisa_layers_attribute: model.layers

wandb_project: CosmoAlpacaLisa-1b-v0.2
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 1
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:

lisa-out

This model is a fine-tuned version of HuggingFaceTB/cosmo-1b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.0525

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
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.2281	0.0	1	1.2636
1.0795	0.25	166	1.0653
1.018	0.5	332	1.0559
1.0363	0.75	498	1.0525

Framework versions

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