See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen1.5-0.5B-Chat
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - b0f300b6b09a2ba8_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/b0f300b6b09a2ba8_train_data.json
  type:
    field_input: my_solu
    field_instruction: prompt
    field_output: solution
    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: 16
gradient_checkpointing: true
group_by_length: false
hub_model_id: Romain-XV/baa021ed-7f27-42f3-8dde-6e74318e8961
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: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_steps: 3726
micro_batch_size: 4
mlflow_experiment_name: /tmp/b0f300b6b09a2ba8_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 1
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: 50fdbd7c-464a-4d56-a3ed-abec091ccab6
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 50fdbd7c-464a-4d56-a3ed-abec091ccab6
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

baa021ed-7f27-42f3-8dde-6e74318e8961

This model is a fine-tuned version of Qwen/Qwen1.5-0.5B-Chat on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1987

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
• training_steps: 3726

Training results

Training Loss	Epoch	Step	Validation Loss
1.3923	0.0002	1	1.4109
0.8141	0.0214	100	0.8870
0.7815	0.0428	200	0.8142
0.7717	0.0642	300	0.7634
0.7092	0.0856	400	0.7206
0.6155	0.1070	500	0.6825
0.7272	0.1284	600	0.6480
0.5709	0.1498	700	0.6154
0.5701	0.1712	800	0.5841
0.574	0.1926	900	0.5533
0.5339	0.2140	1000	0.5240
0.407	0.2354	1100	0.4956
0.4801	0.2568	1200	0.4689
0.4047	0.2783	1300	0.4438
0.4572	0.2997	1400	0.4212
0.3492	0.3211	1500	0.3985
0.3494	0.3425	1600	0.3752
0.38	0.3639	1700	0.3550
0.3508	0.3853	1800	0.3346
0.3265	0.4067	1900	0.3165
0.2892	0.4281	2000	0.3016
0.3141	0.4495	2100	0.2866
0.2954	0.4709	2200	0.2726
0.2968	0.4923	2300	0.2606
0.2706	0.5137	2400	0.2499
0.2689	0.5351	2500	0.2402
0.2072	0.5565	2600	0.2322
0.2147	0.5779	2700	0.2244
0.219	0.5993	2800	0.2175
0.2034	0.6207	2900	0.2129
0.1956	0.6421	3000	0.2093
0.1553	0.6635	3100	0.2053
0.2372	0.6849	3200	0.2028
0.1921	0.7063	3300	0.2009
0.2111	0.7277	3400	0.1996
0.2116	0.7491	3500	0.1990
0.2229	0.7705	3600	0.1987
0.235	0.7920	3700	0.1987

Framework versions

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