lora微调后的glm4模型不生成回答 #4568
Unanswered
RyanCcc114
asked this question in
Q&A
Replies: 2 comments 1 reply
-
overffiting了 參數調整一下,讓學習的特徵不要那麼集中。 |
Beta Was this translation helpful? Give feedback.
1 reply
-
请问这个问题有解决吗。 |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Reminder
System Info
pytorch:2.1.0-cuda11.8
Reproduction
bf16: true
cutoff_len: 1024
dataset: EE_instruction_message
dataset_dir: data
ddp_timeout: 180000000
do_train: true
eval_steps: 100
eval_strategy: steps
finetuning_type: lora
flash_attn: fa2
gradient_accumulation_steps: 8
include_num_input_tokens_seen: true
learning_rate: 0.0003
logging_steps: 5
lora_alpha: 32
lora_dropout: 0.1
lora_rank: 8
lora_target: all
lr_scheduler_type: cosine
max_grad_norm: 1.0
max_samples: 100000
model_name_or_path: src/llamafactory/model/model/glm4-chat
num_train_epochs: 2.0
optim: adamw_torch
output_dir: saves/GLM-4-9B-Chat/lora/train_2024-06-24-23-30-00
packing: false
per_device_eval_batch_size: 1
per_device_train_batch_size: 1
plot_loss: true
preprocessing_num_workers: 16
report_to: none
save_steps: 100
stage: sft
template: glm4
val_size: 0.2
warmup_steps: 0.01
Expected behavior
加载微调后的glm4模型不生成回答,无论输入什么内容,模型返回的内容都是空白。
训练日志中的loss是逐步下降,并且评估阶段的loss同样下降。但是执行predict脚本生成的结果都为0。
Others
No response
Beta Was this translation helpful? Give feedback.
All reactions