See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: jingyeom/seal3.1.6n_7b
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - deb59464c0a667a1_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/deb59464c0a667a1_train_data.json
  type:
    field_instruction: question
    field_output: answer
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 30
eval_max_new_tokens: 128
eval_steps: 50
eval_table_size: null
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 16
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/d1949c4a-0a6f-4ea8-9a91-961471599334
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: true
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
micro_batch_size: 4
mlflow_experiment_name: /tmp/deb59464c0a667a1_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 2
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 100
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 2379ea32-7a7b-48aa-9667-b04a8ab9542e
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 2379ea32-7a7b-48aa-9667-b04a8ab9542e
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

d1949c4a-0a6f-4ea8-9a91-961471599334

This model is a fine-tuned version of jingyeom/seal3.1.6n_7b on the None dataset. It achieves the following results on the evaluation set:

• Loss: nan

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_BNB 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: 2

Training results

Training Loss	Epoch	Step	Validation Loss
0.0	0.0005	1	nan
0.0	0.0273	50	nan
0.0	0.0547	100	nan
0.0	0.0820	150	nan
0.0	0.1093	200	nan
0.0	0.1366	250	nan
0.0	0.1640	300	nan
0.0	0.1913	350	nan
0.0	0.2186	400	nan
0.0	0.2460	450	nan
0.0	0.2733	500	nan
0.0	0.3006	550	nan
0.0	0.3279	600	nan
0.0	0.3553	650	nan
0.0	0.3826	700	nan
0.0	0.4099	750	nan
0.0	0.4372	800	nan
0.0	0.4646	850	nan
0.0	0.4919	900	nan
0.0	0.5192	950	nan
0.0	0.5466	1000	nan
0.0	0.5739	1050	nan
0.0	0.6012	1100	nan
0.0	0.6285	1150	nan
0.0	0.6559	1200	nan
0.0	0.6832	1250	nan
0.0	0.7105	1300	nan
0.0	0.7379	1350	nan
0.0	0.7652	1400	nan
0.0	0.7925	1450	nan
0.0	0.8198	1500	nan
0.0	0.8472	1550	nan
0.0	0.8745	1600	nan

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1