See axolotl config

axolotl version: 0.6.0

base_model: mistralai/Mistral-7B-Instruct-v0.2
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer

trust_remote_code: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: medalpaca/medical_meadow_medqa
    type: alpaca
dataset_prepared_path:
val_set_size: 0.2
output_dir: ./qlora-mistral-7b

sequence_len: 8192
sample_packing: true
eval_sample_packing: true
pad_to_sequence_len: true

adapter: qlora
lora_model_dir:
lora_r: 256
lora_alpha: 128
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 3
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.00002

train_on_inputs: false
group_by_length: false
bf16: true
fp16:
tf32: 

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps:
evals_per_epoch: 4
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
  - full_shard
  - auto_wrap
fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: true
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: true
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_transformer_layer_cls_to_wrap:
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_sharding_strategy: FULL_SHARD
special_tokens:

wandb_project: mistral-7b-instruct-v02
wandb_entity: 
wandb_watch:
wandb_name: 
wandb_log_model: 

hub_model_id: neginashz/mistral-7b-instruct-v02
hub_strategy: 
early_stopping_patience:

resume_from_checkpoint:
auto_resume_from_checkpoints: true
early_stopping_patience:

mistral-7b-instruct-v02

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.2 on the medalpaca/medical_meadow_medqa dataset. It achieves the following results on the evaluation set:

• Loss: 0.1324

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: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 8
• total_eval_batch_size: 8
• optimizer: Use adamw_torch 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: 3
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
0.1395	0.2683	11	0.1455
0.117	0.5366	22	0.1259
0.1221	0.8049	33	0.1206
0.1076	1.0488	44	0.1149
0.0906	1.3171	55	0.1119
0.093	1.5854	66	0.1201
0.0868	1.8537	77	0.1121
0.0634	2.0976	88	0.1146
0.0464	2.3659	99	0.1297
0.0574	2.6341	110	0.1329
0.047	2.9024	121	0.1324

Framework versions

• PEFT 0.14.0
• Transformers 4.47.0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0