See axolotl config

axolotl version: 0.4.1

adapter: qlora
base_model: HuggingFaceH4/zephyr-7b-beta
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - d60bb88622216d48_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/d60bb88622216d48_train_data.json
  type:
    field_instruction: title
    field_output: body
    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: 1
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: error577/b4858ab1-fa33-4d03-bbf8-24158ea20074
hub_repo: null
hub_strategy: end
hub_token: null
learning_rate: 0.0001
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 64
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
max_steps: 500
micro_batch_size: 2
mlflow_experiment_name: /tmp/d60bb88622216d48_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 20
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: 1
sequence_len: 1024
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.002
wandb_entity: null
wandb_mode: online
wandb_name: f4d3b427-02c9-45b2-af1a-2d060051a3e9
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: f4d3b427-02c9-45b2-af1a-2d060051a3e9
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

b4858ab1-fa33-4d03-bbf8-24158ea20074

This model is a fine-tuned version of HuggingFaceH4/zephyr-7b-beta on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.3041

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 500

Training results

Training Loss	Epoch	Step	Validation Loss
10.3685	0.0001	1	2.4300
9.3977	0.0023	25	2.3389
9.9457	0.0046	50	2.3296
9.9118	0.0069	75	2.3246
9.7794	0.0092	100	2.3225
9.5919	0.0115	125	2.3199
9.5492	0.0138	150	2.3179
9.4144	0.0161	175	2.3169
10.2476	0.0184	200	2.3149
9.3078	0.0208	225	2.3130
9.6618	0.0231	250	2.3112
9.01	0.0254	275	2.3107
9.1366	0.0277	300	2.3084
9.4253	0.0300	325	2.3077
9.5503	0.0323	350	2.3070
9.2927	0.0346	375	2.3056
9.6795	0.0369	400	2.3049
9.7989	0.0392	425	2.3045
9.3685	0.0415	450	2.3042
8.9892	0.0438	475	2.3042
7.9958	0.0461	500	2.3041

Framework versions

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