423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-16 11:50:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

[misc] fix packing and eval plot (#7623)

Browse Source
This commit is contained in:
hoshi-hiyouga
2025-04-07 18:20:57 +08:00
committed by GitHub
parent c0079ab9fd
commit fb46193364
70 changed files with 288 additions and 194 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
examples/train_qlora/llama3_lora_sft_aqlm.yaml
View File

@@ -16,6 +16,7 @@ cutoff_len: 2048
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
dataloader_num_workers: 4
### output
output_dir: saves/llama3-8b/lora/sft
@@ -23,6 +24,8 @@ logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none # choices: [none, wandb, tensorboard, swanlab, mlflow]
### train
per_device_train_batch_size: 1

3
examples/train_qlora/llama3_lora_sft_awq.yaml
View File

@@ -16,6 +16,7 @@ cutoff_len: 2048
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
dataloader_num_workers: 4
### output
output_dir: saves/llama3-8b/lora/sft
@@ -23,6 +24,8 @@ logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none # choices: [none, wandb, tensorboard, swanlab, mlflow]
### train
per_device_train_batch_size: 1

5
examples/train_qlora/llama3_lora_sft_bnb_npu.yaml
View File

@@ -1,7 +1,7 @@
### model
model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
quantization_bit: 4
quantization_method: bitsandbytes
quantization_method: bnb
double_quantization: false
trust_remote_code: true
@@ -19,6 +19,7 @@ cutoff_len: 2048
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
dataloader_num_workers: 4
### output
output_dir: saves/llama3-8b/lora/sft
@@ -26,6 +27,8 @@ logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none # choices: [none, wandb, tensorboard, swanlab, mlflow]
### train
per_device_train_batch_size: 1

3
examples/train_qlora/llama3_lora_sft_gptq.yaml
View File

@@ -16,6 +16,7 @@ cutoff_len: 2048
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
dataloader_num_workers: 4
### output
output_dir: saves/llama3-8b/lora/sft
@@ -23,6 +24,8 @@ logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none # choices: [none, wandb, tensorboard, swanlab, mlflow]
### train
per_device_train_batch_size: 1

7
examples/train_qlora/llama3_lora_sft_otfq.yaml
View File

@@ -1,7 +1,7 @@
### model
model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct
quantization_bit: 4
quantization_method: bitsandbytes # choices: [bitsandbytes (4/8), hqq (2/3/4/5/6/8), eetq (8)]
quantization_bit: 4 # choices: [8 (bnb/hqq/eetq), 4 (bnb/hqq), 3 (hqq), 2 (hqq)]
quantization_method: bnb # choices: [bnb, hqq, eetq]
trust_remote_code: true
### method
@@ -18,6 +18,7 @@ cutoff_len: 2048
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
dataloader_num_workers: 4
### output
output_dir: saves/llama3-8b/lora/sft
@@ -25,6 +26,8 @@ logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none # choices: [none, wandb, tensorboard, swanlab, mlflow]
### train
per_device_train_batch_size: 1