You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

Built with Axolotl

See axolotl config

axolotl version: 0.12.0.dev0

base_model: /workspace/data/models/L3.3-Shakudo-70b
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

trust_remote_code: true

load_in_8bit: false
load_in_4bit: true

datasets:
  - path: data/tutu2f_1.03.jsonl
    type: chat_template
    field_messages: messages
    message_property_mappings:
      role: role
      content: content
    roles:
      user: ["user"]
      assistant: ["assistant"]
      system: ["system"]

deepspeed: /workspace/axolotl/deepspeed_configs/zero3.json

dataset_prepared_path: last_run_prepared
output_dir: ./outputs/qlora-omega-shaku-70b
save_steps: 100
save_safetensors: true

adapter: qlora

sequence_len: 5120
sample_packing: false
pad_to_sequence_len: true

lora_r: 32
lora_alpha: 32
lora_dropout: 0.1
lora_target_linear: true

gradient_accumulation_steps: 8
micro_batch_size: 2
num_epochs: 1
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 2e-6
warmup_ratio: 0.05

bf16: auto
tf32: false

gradient_checkpointing: true
logging_steps: 5
flash_attention: true

max_grad_norm: 1.0
weight_decay: 0.0
do_causal_lm_eval: true
eval_steps: 100
eval_causal_lm_metrics: ["sacrebleu", "perplexity"]
special_tokens:
  pad_token: <|finetune_right_pad_id|>

outputs/qlora-omega-shaku-70b

This model was trained from scratch on the data/tutu2f_1.03.jsonl 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: 2e-06
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • distributed_type: multi-GPU
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 16
  • optimizer: Use OptimizerNames.PAGED_ADAMW 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: 63
  • training_steps: 1267

Training results

Framework versions

  • PEFT 0.16.0
  • Transformers 4.54.0
  • Pytorch 2.7.1+cu128
  • Datasets 4.0.0
  • Tokenizers 0.21.2
Downloads last month
-
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for gecfdo/L3.3-The-Omega-Directive-70B-Unslop-v2.1-LoRA