See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: jhflow/mistral7b-lora-multi-turn-v2
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - a8655413fbba9b95_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/a8655413fbba9b95_train_data.json
  type:
    field_input: input
    field_instruction: question
    field_output: answer
    format: '{instruction} {input}'
    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/2720488e-8f03-4ceb-a0aa-55f808e6efac
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/a8655413fbba9b95_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: 48918ade-50c2-4c48-ace4-9c9aafad1b27
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 48918ade-50c2-4c48-ace4-9c9aafad1b27
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

2720488e-8f03-4ceb-a0aa-55f808e6efac

This model is a fine-tuned version of jhflow/mistral7b-lora-multi-turn-v2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7739

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
17.8285	0.0025	1	1.1322
13.9151	0.1263	50	0.8992
14.0505	0.2526	100	0.8687
13.1205	0.3789	150	0.8503
12.4949	0.5052	200	0.8358
13.48	0.6315	250	0.8234
13.5189	0.7578	300	0.8108
12.6055	0.8841	350	0.7991
10.3029	1.0115	400	0.7994
10.7447	1.1378	450	0.7965
10.5841	1.2641	500	0.7910
9.9121	1.3904	550	0.7827
9.5545	1.5167	600	0.7795
9.9359	1.6430	650	0.7771
10.6437	1.7693	700	0.7737
9.6907	1.8956	750	0.7739

Framework versions

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