See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: bigcode/starcoder2-3b
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 747ca939a112ba35_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/747ca939a112ba35_train_data.json
  type:
    field_instruction: instruction
    field_output: output
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 4
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
gradient_clipping: 1.0
group_by_length: false
hub_model_id: leixa/8d76f321-85fe-406c-b985-e6e122a1f424
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 5.0e-05
load_in_4bit: false
load_in_8bit: false
local_rank: 0
logging_steps: 3
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_steps: 150
micro_batch_size: 8
mlflow_experiment_name: /tmp/747ca939a112ba35_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
saves_per_epoch: 4
sequence_len: 1024
special_tokens:
  pad_token: <|endoftext|>
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: techspear-hub
wandb_mode: online
wandb_name: bdde0a17-2182-4c18-855c-68558222f915
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: bdde0a17-2182-4c18-855c-68558222f915
warmup_steps: 10
weight_decay: 0.01
xformers_attention: null

8d76f321-85fe-406c-b985-e6e122a1f424

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: 1.0542

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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
• training_steps: 150

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0006	1	1.0594
16.1021	0.0074	13	1.0563
11.192	0.0149	26	1.0759
8.986	0.0223	39	1.0989
8.4314	0.0298	52	1.0748
7.2741	0.0372	65	1.0764
7.4998	0.0447	78	1.0654
6.8855	0.0521	91	1.0637
7.4119	0.0596	104	1.0567
6.8738	0.0670	117	1.0547
6.7777	0.0744	130	1.0546
6.8785	0.0819	143	1.0542

Framework versions

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