Hugging Face's logo

afrias5
/
meta-codellama-34b-python-Score8192V4

PEFT
Safetensors
llama
axolotl
Generated from Trainer
Model card Files Community
meta-codellama-34b-python-Score8192V4 / README.md
afrias5's picture
afrias5
Model save
bafaa1c verified
preview code
|
raw
history blame contribute delete
3.42 kB
metadata
library_name: peft
license: llama2
base_model: meta-llama/CodeLlama-34b-Python-hf
tags:
  - axolotl
  - generated_from_trainer
datasets:
  - afrias5/datasetScoreFinal
model-index:
  - name: meta-codellama-34b-python-Score8192V4
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.5.3.dev41+g5e9fa33f

base_model: meta-llama/CodeLlama-34b-Python-hf
model_type: LlamaForCausalLM
tokenizer_type: CodeLlamaTokenizer


load_in_8bit: false
load_in_4bit: false
strict: false


datasets:
  - path: afrias5/datasetScoreFinal
    type: alpaca
    field: text


# dataset_prepared_path: ./FinUpTagsNoTestNoExNew
val_set_size: 0
output_dir: models/meta-codellama-34b-python-Score8192V4
lora_model_dir: models/meta-codellama-34b-python-Score8192V4/checkpoint-55
auto_resume_from_checkpoints: true
sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: False
adapter: lora
lora_model_dir:
lora_r: 2
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_modules_to_save:
  - embed_tokens
  - lm_head


wandb_project: 'Code34bNewFeed'
wandb_entity:
wandb_watch:
wandb_run_id:
wandb_name: 'meta-codellama-34b-python-Score8192V4'                            
wandb_log_model:


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


train_on_inputs: false
group_by_length: false
bf16: true
fp16:
tf32: false
hub_model_id: afrias5/meta-codellama-34b-python-Score8192V4
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: false
s2_attention:
logging_steps: 1
warmup_steps: 10
saves_per_epoch: 1
save_total_limit: 16
debug:
deepspeed:
weight_decay: 0.0
fsdp:
deepspeed: deepspeed_configs/zero3_bf16_cpuoffload_all.json
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

meta-codellama-34b-python-Score8192V4

This model is a fine-tuned version of meta-llama/CodeLlama-34b-Python-hf on the afrias5/datasetScoreFinal dataset.

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: 1
  • eval_batch_size: 1
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 2
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 8
  • total_eval_batch_size: 2
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • num_epochs: 14

Training results

Framework versions

  • PEFT 0.14.0
  • Transformers 4.46.3
  • Pytorch 2.5.1+cu121
  • Datasets 3.1.0
  • Tokenizers 0.20.3