CognitiveLab/Ambari-One-Thrid-PEFT_final

See axolotl config

axolotl version: 0.4.0

base_model: meta-llama/Llama-2-7b-hf
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer
tokenizer_config: meta-llama/Llama-2-7b-hf
is_llama_derived_model: true

load_in_8bit: true
load_in_4bit: false
strict: false

datasets:
  - path: Cognitive-Lab/Kannada_Bilingual_Instruct
    type: completion
dataset_prepared_path: 
val_set_size: 0.05
output_dir: ./kannada_out_peft_instruct_final_no_ext

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 256
lora_alpha: 128 
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project: Ambari-Instruct-No-Extension
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 14
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: false
bf16: true
fp16: false
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: 1
eval_table_size:
eval_table_max_new_tokens: 128
saves_per_epoch: 10
debug:
# deepspeed: deepspeed/zero2.json
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

kannada_out_peft_instruct_final_no_ext

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

• Loss: 0.4334

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: 14
• eval_batch_size: 14
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 112
• total_eval_batch_size: 28
• 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
0.3109	1.0	888	0.4334

Framework versions

• PEFT 0.9.0
• Transformers 4.40.0.dev0
• Pytorch 2.1.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.0