base_model: huihui-ai/Llama-3.3-70B-Instruct-abliterated-finetuned
load_in_8bit: false
load_in_4bit: true
adapter: qlora
wandb_name: r1_axolotl_ft
output_dir: ./outputs/out/r1_axolotl_ft
hub_model_id: cgifbribcgfbi/Llama-3.3-70B-Instruct-abliterated-finetuned-chemistry-r1
hub_strategy: every_save
# resume_from_checkpoint: ./outputs/out/5_70B_axolotl_ft/checkpoint-72

tokenizer_type: AutoTokenizer
push_dataset_to_hub:
strict: false

datasets:
  - path: dset_r1_clean_4998.jsonl
    type: chat_template
    split: train

dataset_prepared_path: last_run_prepared
val_set_size: 0.05
# test_datasets:
#   - path: 5000_benign_val.json
#     type: chat_template
#     split: train
save_safetensors: true

sequence_len: 3500
sample_packing: true
pad_to_sequence_len: true

lora_r: 64
lora_alpha: 32
lora_dropout: 0.05
lora_target_modules:
lora_target_linear: true

wandb_mode:
wandb_project: finetune-chem
wandb_entity: gpoisjgqetpadsfke
wandb_watch:
wandb_run_id:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 4
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 0.00002

train_on_inputs: false
group_by_length: true
bf16: true
tf32: true

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: true
logging_steps: 1
flash_attention: true

warmup_steps: 10
evals_per_epoch: 3
saves_per_epoch: 1
weight_decay: 0.01
fsdp:
  - full_shard
  - auto_wrap
fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: false
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: true
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_sharding_strategy: FULL_SHARD
special_tokens:
  pad_token: <|finetune_right_pad_id|>

Llama-3.3-70B-Instruct-abliterated-finetuned-chemistry-r1

This model is a fine-tuned version of huihui-ai/Llama-3.3-70B-Instruct-abliterated-finetuned on the dset_r1_clean_4998.jsonl dataset. It achieves the following results on the evaluation set:

Loss: 0.3558

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-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 4
total_train_batch_size: 8
total_eval_batch_size: 8
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 4.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.7163	0.0037	1	0.7191
0.4295	0.3333	90	0.4328
0.4114	0.6667	180	0.3992
0.382	1.0	270	0.3827
0.3651	1.3333	360	0.3736
0.3614	1.6667	450	0.3679
0.3644	2.0	540	0.3630
0.3201	2.3333	630	0.3608
0.3179	2.6667	720	0.3580
0.3103	3.0	810	0.3561
0.3043	3.3333	900	0.3562
0.3098	3.6667	990	0.3558
0.2853	4.0	1080	0.3558

Framework versions

PEFT 0.15.1
Transformers 4.51.0
Pytorch 2.5.1+cu124
Datasets 3.5.0
Tokenizers 0.21.1