See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: fxmarty/tiny-dummy-qwen2
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 50c9e2f5e890969e_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/50c9e2f5e890969e_train_data.json
  type:
    field_instruction: instruction
    field_output: chosen_response
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
device_map:
  ? ''
  : 0,1,2,3,4,5,6,7
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/dfe5b386-7689-4b46-8540-dc47ace2edfb
hub_repo: null
hub_strategy: null
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: 128
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 4679
micro_batch_size: 4
mlflow_experiment_name: /tmp/50c9e2f5e890969e_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: ac700e19-5242-448c-86d7-7306a9164c1c
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: ac700e19-5242-448c-86d7-7306a9164c1c
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

dfe5b386-7689-4b46-8540-dc47ace2edfb

This model is a fine-tuned version of fxmarty/tiny-dummy-qwen2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 11.9163

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: 4109

Training results

Training Loss	Epoch	Step	Validation Loss
11.9329	0.0005	1	11.9324
11.9218	0.0487	100	11.9209
11.9212	0.0974	200	11.9200
11.9206	0.1460	300	11.9193
11.9165	0.1947	400	11.9187
11.9205	0.2434	500	11.9182
11.9184	0.2921	600	11.9180
11.9192	0.3407	700	11.9178
11.9175	0.3894	800	11.9176
11.9191	0.4381	900	11.9174
11.9165	0.4868	1000	11.9174
11.9178	0.5354	1100	11.9172
11.9183	0.5841	1200	11.9172
11.9185	0.6328	1300	11.9170
11.9162	0.6815	1400	11.9170
11.918	0.7301	1500	11.9168
11.9175	0.7788	1600	11.9169
11.9171	0.8275	1700	11.9167
11.9182	0.8762	1800	11.9167
11.9166	0.9249	1900	11.9166
11.9176	0.9735	2000	11.9166
12.3114	1.0222	2100	11.9165
13.4333	1.0709	2200	11.9166
15.1108	1.1196	2300	11.9165
12.4772	1.1682	2400	11.9165
12.2957	1.2169	2500	11.9164
12.2061	1.2656	2600	11.9164
13.2965	1.3143	2700	11.9164
12.2792	1.3629	2800	11.9164
11.7337	1.4116	2900	11.9163
11.2628	1.4603	3000	11.9163
12.6356	1.5090	3100	11.9163
11.6645	1.5577	3200	11.9163
12.7076	1.6063	3300	11.9163
12.6979	1.6550	3400	11.9163
12.171	1.7037	3500	11.9163

Framework versions

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