mhhmm's picture
Upload 18 files
4396cbf verified
base_model: codellama/CodeLlama-13b-Instruct-hf
model_type: LlamaForCausalLM
tokenizer_type: CodeLlamaTokenizer
is_llama_derived_model: true
load_in_8bit: false
bf16: true
strict: false
datasets:
- path: data.jsonl
ds_type: json
type:
# JSONL file contains question, context, answer fields per line.
# This gets mapped to instruction, input, output axolotl tags.
field_instruction: instruction
#field_input: context
field_output: output
# Format is used by axolotl to generate the prompt.
format: |-
Using the instruction context below, generate a typescript code that answers the question and explain it
{instruction}
dataset_prepared_path:
val_set_size: 16 # must be at least micro_batch_size * N_GPUS, and more if eval packing.
output_dir: ./lora-out
sequence_len: 4096
sample_packing: true
eval_sample_packing: false
pad_to_sequence_len: true
adapter: lora
lora_model_dir:
lora_r: 16
lora_alpha: 32 # alpha = 2 x rank is a good starting point.
lora_dropout: 0.05
lora_target_linear: true # target all linear layers
lora_fan_in_fan_out:
wandb_project:
wandb_entity:
wandb_watch:
wandb_run_id:
gradient_accumulation_steps: 1
micro_batch_size: 8
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
train_on_inputs: false
group_by_length: false
fp16: false
tf32: false
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
auto_resume_from_checkpoints: true
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
warmup_steps: 10
eval_steps: 0.05
save_steps:
debug: True
deepspeed: /root/axolotl/deepspeed/zero3.json
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
bos_token: "<s>"
eos_token: "</s>"
unk_token: "<unk>"