See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.1-Storm-8B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - b51894bc358e69ef_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/b51894bc358e69ef_train_data.json
  type:
    field_instruction: question_en
    field_output: chosen_en
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 30
eval_max_new_tokens: 128
eval_steps: 50
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 16
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/8b6619f1-e8e2-4475-a003-33211e86f690
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
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: true
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
micro_batch_size: 4
mlflow_experiment_name: /tmp/b51894bc358e69ef_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: 2048
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: 80865761-d38a-4c48-9ea1-fb8848e26565
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 80865761-d38a-4c48-9ea1-fb8848e26565
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

8b6619f1-e8e2-4475-a003-33211e86f690

This model is a fine-tuned version of unsloth/Llama-3.1-Storm-8B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.1177

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: 16
• total_train_batch_size: 64
• 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: 2

Training results

Training Loss	Epoch	Step	Validation Loss
2.0186	0.0008	1	2.0494
1.4436	0.0406	50	1.3902
1.2774	0.0812	100	1.3402
1.3279	0.1218	150	1.3092
1.2515	0.1625	200	1.2914
1.2672	0.2031	250	1.2785
1.2416	0.2437	300	1.2647
1.2159	0.2843	350	1.2555
1.1702	0.3249	400	1.2434
1.2178	0.3655	450	1.2364
1.3226	0.4062	500	1.2286
1.1198	0.4468	550	1.2208
1.2484	0.4874	600	1.2148
1.1048	0.5280	650	1.2099
1.2055	0.5686	700	1.2041
1.219	0.6092	750	1.1982
1.2205	0.6498	800	1.1918
1.275	0.6905	850	1.1856
1.1738	0.7311	900	1.1803
1.1674	0.7717	950	1.1752
1.1784	0.8123	1000	1.1705
1.1904	0.8529	1050	1.1673
1.1293	0.8935	1100	1.1626
1.1344	0.9342	1150	1.1571
1.1763	0.9748	1200	1.1534
1.0566	1.0154	1250	1.1584
1.0529	1.0560	1300	1.1588
1.1357	1.0966	1350	1.1547
1.0739	1.1372	1400	1.1519
0.9531	1.1778	1450	1.1495
1.0049	1.2185	1500	1.1456
0.9675	1.2591	1550	1.1431
0.9961	1.2997	1600	1.1408
0.9481	1.3403	1650	1.1355
0.9777	1.3809	1700	1.1339
1.031	1.4215	1750	1.1327
1.0313	1.4622	1800	1.1315
0.932	1.5028	1850	1.1272
1.003	1.5434	1900	1.1260
1.0032	1.5840	1950	1.1236
0.913	1.6246	2000	1.1231
0.9941	1.6652	2050	1.1222
0.9415	1.7058	2100	1.1204
1.0222	1.7465	2150	1.1188
0.9803	1.7871	2200	1.1183
0.9454	1.8277	2250	1.1181
0.988	1.8683	2300	1.1182
0.9538	1.9089	2350	1.1179
0.9843	1.9495	2400	1.1179
0.8832	1.9902	2450	1.1177

Framework versions

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