See axolotl config

axolotl version: 0.4.1

adapter: qlora
base_model: unsloth/Qwen2-1.5B
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 0c4e37d6699a2850_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/0c4e37d6699a2850_train_data.json
  type:
    field_input: multi_turn_queries
    field_instruction: actor_name
    field_output: plain_query
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 1
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: error577/55d263a8-fd6f-4708-81c2-9aef4ad36745
hub_repo: null
hub_strategy: end
hub_token: null
learning_rate: 0.0001
load_in_4bit: true
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: 32
lora_target_linear: true
lr_scheduler: cosine
max_steps: 500
micro_batch_size: 1
mlflow_experiment_name: /tmp/0c4e37d6699a2850_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 20
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
saves_per_epoch: 1
sequence_len: 512
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.02
wandb_entity: null
wandb_mode: online
wandb_name: 250041e6-2183-4380-94e2-2d0d237a106e
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 250041e6-2183-4380-94e2-2d0d237a106e
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

55d263a8-fd6f-4708-81c2-9aef4ad36745

This model is a fine-tuned version of unsloth/Qwen2-1.5B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.7902

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: 0.0001
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• 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: 500

Training results

Training Loss	Epoch	Step	Validation Loss
3.9966	0.0055	1	4.2648
2.7104	0.1379	25	2.7792
2.3145	0.2759	50	2.2709
2.0065	0.4138	75	2.2007
2.0039	0.5517	100	2.0560
2.1108	0.6897	125	1.9981
1.6955	0.8276	150	1.9705
1.7898	0.9655	175	1.9222
0.8951	1.1034	200	1.8972
2.0605	1.2414	225	1.8945
1.3584	1.3793	250	1.8671
2.7196	1.5172	275	1.8422
1.4262	1.6552	300	1.8102
1.1387	1.7931	325	1.7863
1.3279	1.9310	350	1.7713
1.1366	2.0690	375	1.7653
0.7406	2.2069	400	1.7976
1.1988	2.3448	425	1.7904
1.3817	2.4828	450	1.7899
0.8481	2.6207	475	1.7891
1.5148	2.7586	500	1.7902

Framework versions

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