Qwen1.5-8x7b-v0.1 / README.md
MaziyarPanahi's picture
7fd6eff2e7ab7e53b7793b849bd6f7465bb19239d732882a8b38a23b51467d97
b4742a2 verified
|
raw
history blame
3.26 kB
metadata
library_name: peft
tags:
  - axolotl
  - generated_from_trainer
  - moe
  - qwen
  - text-generation-inference
base_model: MaziyarPanahi/Qwen1.5-8x7b
model-index:
  - name: Qwen1.5-8x7b-v0.1
    results: []

Built with Axolotl

See axolotl config

axolotl version: 0.4.0

base_model: MaziyarPanahi/Qwen1.5-8x7b
model_type: Qwen2ForCausalLM
tokenizer_type: Qwen2Tokenizer

trust_remote_code: true

hub_model_id: MaziyarPanahi/Qwen1.5-8x7b-v0.1
hf_use_auth_token: true

load_in_8bit: false
load_in_4bit: true
strict: false


datasets:
  - path: Crystalcareai/MoD-150k
    type: sharegpt


dataset_prepared_path:
val_set_size: 0.05
output_dir: ./Qwen1.5-8x7b-v0.1-lora-out

model_config:
  output_router_logits: true

adapter: qlora
lora_model_dir:
sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true


lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:


gradient_accumulation_steps: 2
micro_batch_size: 2
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002


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


warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

Qwen1.5-8x7b-v0.1

This model is a fine-tuned version of MaziyarPanahi/Qwen1.5-8x7b on the Crystalcareai/MoD-150k dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7945

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: 2
  • eval_batch_size: 2
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 4
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • total_eval_batch_size: 8
  • 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
6.2196 0.0 1 6.1942
0.7772 0.25 513 0.8037
0.656 0.5 1026 0.7977
0.6967 0.75 1539 0.7945

Framework versions

  • PEFT 0.8.2
  • Transformers 4.39.0.dev0
  • Pytorch 2.2.0+cu121
  • Datasets 2.17.0
  • Tokenizers 0.15.0