See axolotl config

axolotl version: 0.4.1

adapter: qlora
auto_resume_from_checkpoints: true
base_model: bigcode/starcoder2-3b
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - f2d6636a6bf983c9_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/f2d6636a6bf983c9_train_data.json
  type:
    field_instruction: formal_statement
    field_output: formal_proof
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 3
eval_max_new_tokens: 128
eval_steps: 50
eval_table_size: null
evals_per_epoch: null
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 16
gradient_checkpointing: true
group_by_length: true
hub_model_id: error577/7b6962a8-57ff-4af2-a962-f8a0573ee5c1
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0003
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 128
lora_dropout: 0.3
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 128
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: null
micro_batch_size: 1
mlflow_experiment_name: /tmp/f2d6636a6bf983c9_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
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: 50
sequence_len: 512
special_tokens:
  pad_token: <|endoftext|>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.02
wandb_entity: null
wandb_mode: online
wandb_name: ff38011c-5505-4460-be03-6a4b9792853e
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: ff38011c-5505-4460-be03-6a4b9792853e
warmup_steps: 10
weight_decay: 0.01
xformers_attention: null

7b6962a8-57ff-4af2-a962-f8a0573ee5c1

This model is a fine-tuned version of bigcode/starcoder2-3b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5802

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: 0.0003
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 16
• 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: 3

Training results

Training Loss	Epoch	Step	Validation Loss
88.6049	0.0006	1	2.9551
21.7947	0.0299	50	0.9747
18.2951	0.0598	100	0.8145
14.9501	0.0897	150	0.8101
12.5567	0.1196	200	0.7215
11.8721	0.1495	250	0.7550
12.4923	0.1794	300	0.7475
11.8208	0.2093	350	0.6632
10.4634	0.2392	400	0.6835
12.3047	0.2691	450	0.6547
17.5321	0.2990	500	0.6600
9.612	0.3289	550	0.6275
8.8472	0.3588	600	0.6169
7.3696	0.3887	650	0.5939
11.7025	0.4186	700	0.5968
7.1958	0.4485	750	0.5639
12.7003	0.4784	800	0.5736
8.3969	0.5083	850	0.5705
12.0703	0.5382	900	0.5802

Framework versions

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