Howard881010/climate_numerical

See axolotl config

axolotl version: 0.4.1

base_model: meta-llama/Llama-2-7b-chat-hf
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: Howard881010/climate-numerical
    type: alpaca
dataset_prepared_path:
val_set_size: 0.05
output_dir: ./finetune/outputs/climate-numerical-finetune

adapter: qlora
lora_model_dir:

sequence_len: 170
sample_packing: false
pad_to_sequence_len: true

lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_modules:
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project: finetune
wandb_entity:
wandb_watch:
wandb_name: climate-numerical-finetune
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 4
optimizer: paged_adamw_32bit
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
eval_sample_packing: False

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

finetune/outputs/climate-numerical-finetune

This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7363

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: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 16
• 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
1.3514	0.0889	1	1.3063
1.3508	0.2667	3	1.2867
1.1945	0.5333	6	1.0302
0.8568	0.8	9	0.8416
0.8142	1.0667	12	0.7761
0.7736	1.3333	15	0.7398
0.7616	1.6	18	0.7341
0.7554	1.8667	21	0.7368
0.7499	2.1333	24	0.7389
0.737	2.4	27	0.7380
0.7397	2.6667	30	0.7364
0.7284	2.9333	33	0.7366
0.725	3.2	36	0.7368
0.7405	3.4667	39	0.7365
0.7315	3.7333	42	0.7363

Framework versions

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