See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2.5-3B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - bdd8a35f55f25533_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/bdd8a35f55f25533_train_data.json
  type:
    field_input: original_version
    field_instruction: title
    field_output: french_version
    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: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/8d2e89e7-9f4b-4702-ae4c-8d1cc863cb26
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.00025
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 64
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 2760
micro_batch_size: 4
mlflow_experiment_name: /tmp/bdd8a35f55f25533_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
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: 117bcf8a-89aa-4e58-88c4-fd9dde22f122
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 117bcf8a-89aa-4e58-88c4-fd9dde22f122
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

8d2e89e7-9f4b-4702-ae4c-8d1cc863cb26

This model is a fine-tuned version of unsloth/Qwen2.5-3B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8278

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: 0.00025
• 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: 2760

Training results

Training Loss	Epoch	Step	Validation Loss
1.0386	0.0003	1	1.1236
1.142	0.0343	100	0.9476
1.1331	0.0686	200	0.9250
1.0164	0.1029	300	0.9124
0.7867	0.1372	400	0.9005
0.9931	0.1715	500	0.8936
0.9221	0.2058	600	0.8876
0.9674	0.2401	700	0.8808
0.8076	0.2744	800	0.8757
1.2683	0.3087	900	0.8702
1.0743	0.3431	1000	0.8658
1.0382	0.3774	1100	0.8619
0.8643	0.4117	1200	0.8574
0.7861	0.4460	1300	0.8534
0.8088	0.4803	1400	0.8501
0.802	0.5146	1500	0.8461
1.159	0.5489	1600	0.8436
0.8837	0.5832	1700	0.8403
1.0334	0.6175	1800	0.8375
0.9367	0.6518	1900	0.8353
1.0622	0.6861	2000	0.8333
0.9135	0.7204	2100	0.8319
0.8457	0.7547	2200	0.8304
0.6793	0.7890	2300	0.8294
0.8514	0.8233	2400	0.8286
0.8636	0.8576	2500	0.8282
0.9369	0.8919	2600	0.8279
0.9562	0.9262	2700	0.8278

Framework versions

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