Hugging Face's logo

liuylhf
/
mixtral-lora

PEFT
Safetensors
mixtral
axolotl
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Community
mixtral-lora / README.md
liuylhf's picture
liuylhf
End of training
b258909 verified
preview code
|
raw
history blame
4.08 kB
metadata
license: apache-2.0
library_name: peft
tags:
  - axolotl
  - generated_from_trainer
base_model: mistralai/Mixtral-8x7B-Instruct-v0.1
model-index:
  - name: mixtral-lora
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.4.0

base_model: mistralai/Mixtral-8x7B-Instruct-v0.1
model_type: AutoModelForCausalLM
tokenizer_type: LlamaTokenizer
is_mistral_derived_model: true
load_in_8bit: false
load_in_4bit: true
strict: false
chat_template: inst

datasets:
datasets:
  - path: ./data/raw_format/tool_used_training.jsonl
    type: sharegpt
  - path: ./data/raw_format/tool_not_used_training.jsonl
    type: sharegpt
  - path: ./data/raw_format/no_tools_training.jsonl
    type: sharegpt

dataset_prepared_path: last_run_prepared
val_set_size: 0.01
output_dir: ../../text-generation-webui/loras/mixtral-instruct-raw-data-v3-inst

adapter: lora
lora_model_dir:

sequence_len: 4096
sample_packing: true
pad_to_sequence_len: true

lora_r: 16
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
hub_model_id: liuylhf/mixtral-lora

# lora_target_modules:
#   - gate_proj
#   - down_proj
#   - up_proj
#   - q_proj
#   - v_proj
#   - k_proj
#   - o_proj

wandb_project: function-call
wandb_name: mixtral-instruct-raw-data-v3
wandb_log_model: end

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 2
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 0.001
adam_beta2: 0.95
adam_epsilon: 0.00001
max_grad_norm: 1.0

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

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

# loss_watchdog_threshold: 5.0
# loss_watchdog_patience: 3

warmup_steps: 10
# evals_per_epoch: 20
eval_steps: 0.1
save_steps: 0.1
eval_table_size:
eval_max_new_tokens: 256
# saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 1.0
fsdp:
fsdp_config:

mixtral-lora

This model is a fine-tuned version of mistralai/Mixtral-8x7B-Instruct-v0.1 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1746

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
3.397 0.0 1 3.2822
0.1294 0.2 67 0.2029
0.1664 0.4 134 0.1918
0.1742 0.6 201 0.1853
0.163 0.8 268 0.1827
0.1537 1.0 335 0.1798
0.1056 1.19 402 0.1781
0.1688 1.39 469 0.1765
0.1187 1.59 536 0.1752
0.1823 1.79 603 0.1748
0.1022 1.99 670 0.1746

Framework versions

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