shuttie/mistral-7b-dadjokes-v2

See axolotl config

axolotl version: 0.4.1

base_model: mistralai/Mistral-7B-v0.3
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false
val_set_size: 0.01
datasets:
  - path: shuttie/reddit-dadjokes
    split: train
    type: alpaca

dataset_prepared_path: last_run_prepared
output_dir: ./outputs/dadjoke-mistral-qlora-out

adapter: qlora
lora_model_dir:

sequence_len: 256
sample_packing: false
pad_to_sequence_len: true

lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_modules:
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: 60
num_epochs: 1
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.00005

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
xformers_attention:
flash_attention: true

logging_steps: 10
warmup_steps: 10
evals_per_epoch: 10
eval_table_size:
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: false
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: false
  fsdp_transformer_layer_cls_to_wrap: MistralDecoderLayer
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
special_tokens:
# torch_compile: true
# chat_template: chatml

outputs/dadjoke-mistral-qlora-out

This model is a fine-tuned version of mistralai/Mistral-7B-v0.3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.2797

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: 5e-05
• train_batch_size: 60
• eval_batch_size: 60
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 120
• total_eval_batch_size: 120
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0008	1	2.9205
2.3515	0.1001	122	2.3554
2.2695	0.2002	244	2.3219
2.3065	0.3002	366	2.3112
2.2109	0.4003	488	2.2974
2.2043	0.5004	610	2.2941
2.2672	0.6005	732	2.2878
2.2259	0.7006	854	2.2825
2.2386	0.8007	976	2.2820
2.247	0.9007	1098	2.2797

Framework versions

• PEFT 0.11.1
• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1