See axolotl config

axolotl version: 0.4.0

adam_beta2: 0.95
adam_epsilon: 1.0e-05
adapter: lora
base_model: mistralai/Mistral-7B-Instruct-v0.2
bf16: auto
chat_template: inst
dataset_prepared_path: last_run_prepared
datasets:
- conversation: mistral
  path: 4e9501d816a24795b7d619faea6fe0b7/./data/raw_format/tool_used_training_small.jsonl
  type: sharegpt
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 256
eval_steps: 0.2
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: liuylhf/mistral-lora
is_mistral_derived_model: true
learning_rate: 0.001
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
micro_batch_size: 2
model_type: AutoModelForCausalLM
num_epochs: 2
optimizer: paged_adamw_8bit
output_dir: ../../text-generation-webui/loras/mistral-instruct-raw-format-v2-more-positive-inst
pad_to_sequence_len: true
resume_from_checkpoint: null
sample_packing: true
save_steps: 0.2
sequence_len: 4096
strict: false
tf32: false
tokenizer_type: LlamaTokenizer
train_on_inputs: false
val_set_size: 0.1
wandb_log_model: end
wandb_name: mixtral-instruct-qlora-v1
wandb_project: function-call
warmup_steps: 10
weight_decay: 1.0
xformers_attention: null

mistral-lora

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0298

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: 0.001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss
2.2964	0.02	1	2.1559
0.0487	0.41	21	0.0479
0.0367	0.81	42	0.0387
0.0331	1.19	63	0.0333
0.0209	1.6	84	0.0298

Framework versions

• PEFT 0.8.2
• Transformers 4.39.0.dev0
• Pytorch 2.2.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.0