See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen1.5-1.8B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - b9b47b46d5c01cab_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/b9b47b46d5c01cab_train_data.json
  type:
    field_input: source
    field_instruction: title
    field_output: comment
    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: 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/b82d5463-6b89-4a7b-bbb2-1c4760d1f05b
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: true
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
lr_scheduler: cosine
max_steps: 518
micro_batch_size: 4
mlflow_experiment_name: /tmp/b9b47b46d5c01cab_train_data.json
model_type: AutoModelForCausalLM
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: e3fe0c30-ee99-44d8-983f-a54143508d00
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: e3fe0c30-ee99-44d8-983f-a54143508d00
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

b82d5463-6b89-4a7b-bbb2-1c4760d1f05b

This model is a fine-tuned version of Qwen/Qwen1.5-1.8B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.8768

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: 518

Training results

Training Loss	Epoch	Step	Validation Loss
5.2371	0.0017	1	5.1936
4.3345	0.0837	50	4.2566
4.1078	0.1675	100	4.1231
4.1016	0.2512	150	4.0534
3.8452	0.3349	200	4.0041
3.9826	0.4186	250	3.9643
4.0881	0.5024	300	3.9292
3.6798	0.5861	350	3.9057
3.9747	0.6698	400	3.8880
3.958	0.7535	450	3.8795
3.8107	0.8373	500	3.8768

Framework versions

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