See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: bigscience/bloomz-560m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 25381d550887dd6e_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/25381d550887dd6e_train_data.json
  type:
    field_input: operators
    field_instruction: question_text
    field_output: decomposition
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: false
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/a8b12708-2aae-43c9-9def-7df8e7127f74
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
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: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules: null
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 6528
micro_batch_size: 4
mlflow_experiment_name: /tmp/25381d550887dd6e_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: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: d7f41241-8e43-47ec-9de5-2ca5bb717ff5
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: d7f41241-8e43-47ec-9de5-2ca5bb717ff5
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

a8b12708-2aae-43c9-9def-7df8e7127f74

This model is a fine-tuned version of bigscience/bloomz-560m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2028

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: 8
• 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: 3591

Training results

Training Loss	Epoch	Step	Validation Loss
39.0838	0.0006	1	5.0617
5.0744	0.0557	100	0.5737
4.0586	0.1114	200	0.4370
3.4828	0.1671	300	0.3907
3.6622	0.2228	400	0.3530
2.7543	0.2785	500	0.3289
2.1382	0.3342	600	0.3228
2.8589	0.3899	700	0.3082
2.5821	0.4456	800	0.2919
1.8472	0.5013	900	0.2815
1.7843	0.5569	1000	0.2699
2.0657	0.6126	1100	0.2706
1.9569	0.6683	1200	0.2613
2.3046	0.7240	1300	0.2549
2.2584	0.7797	1400	0.2501
1.9766	0.8354	1500	0.2448
1.8655	0.8911	1600	0.2389
1.881	0.9468	1700	0.2374
1.5574	1.0025	1800	0.2298
1.422	1.0582	1900	0.2268
1.7154	1.1139	2000	0.2238
1.5083	1.1696	2100	0.2252
1.4274	1.2253	2200	0.2204
1.8438	1.2810	2300	0.2159
1.72	1.3367	2400	0.2144
1.4157	1.3924	2500	0.2137
1.6784	1.4481	2600	0.2107
1.3869	1.5038	2700	0.2092
1.4246	1.5595	2800	0.2077
1.4807	1.6151	2900	0.2070
1.0326	1.6708	3000	0.2067
1.6354	1.7265	3100	0.2044
1.4867	1.7822	3200	0.2049
1.1222	1.8379	3300	0.2035
1.6541	1.8936	3400	0.2048
1.2433	1.9493	3500	0.2028

Framework versions

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