ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update llama.py

Browse source
This commit is contained in:
Daniel Han 2025-03-03 23:32:02 -08:00
commit 69044ce605
1 changed files with 7 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

14
unsloth/models/llama.py
View file

@ -1684,10 +1684,10 @@ class FastLlamaModel:
statistics = \
f"==((====))== Unsloth {__version__}: Fast {model_patcher.__name__[4:-5]} patching. Transformers: {transformers_version}.\n"\
f" {chr(92)}{chr(92)} /| GPU: {gpu_stats.name}. Max memory: {max_memory} GB. Platform: {platform_system}.\n"\
f" {chr(92)}{chr(92)} /| {gpu_stats.name}. Num GPUs = {torch.cuda.device_count()}. Max memory: {max_memory} GB. Platform: {platform_system}.\n"\
f"O^O/ {chr(92)}_/ {chr(92)} Torch: {torch.__version__}. CUDA: {gpu_stats.major}.{gpu_stats.minor}. CUDA Toolkit: {torch.version.cuda}. Triton: {triton_version}\n"\
f"{chr(92)} / Bfloat16 = {str(SUPPORTS_BFLOAT16).upper()}. FA [Xformers = {xformers_version}. FA2 = {HAS_FLASH_ATTENTION}]\n"\
f' "-____-" Free Apache license: http://github.com/unslothai/unsloth'
f' "-____-" Free license: http://github.com/unslothai/unsloth'
print(statistics)
# Warn about fast transfers
@ -1879,11 +1879,11 @@ class FastLlamaModel:
# Cannot use \\ since it will cause a SyntaxWarning in Python 3.12
# Instead use chr(92) == \\
debug_info = """debug_info = \\
f"==((====))== Unsloth - 2x faster free finetuning | Num GPUs = {args.world_size}\\n"\\
f" {chr(92)}{chr(92)} /| Num examples = {num_examples:,} | Num Epochs = {num_train_epochs:,}\\n"\\
f"O^O/ {chr(92)}_/ {chr(92)} Batch size per device = {self._train_batch_size:,} | Gradient Accumulation steps = {args.gradient_accumulation_steps}\\n"\\
f"{chr(92)} / Total batch size = {total_train_batch_size:,} | Total steps = {max_steps:,}\\n"\\
f' "-____-" Number of trainable parameters = {get_model_param_count(model, trainable_only=True):,}'
f"==((====))== Unsloth - 2x faster free finetuning | Num GPUs used = {len(set(p.device for p in model.parameters()))}\\n"\\
f" {chr(92)}{chr(92)} /| Num examples = {num_examples:,} | Num Epochs = {num_train_epochs:,} | Total steps = {max_steps:,}\\n"\\
f"O^O/ {chr(92)}_/ {chr(92)} Batch size per device = {self._train_batch_size:,} | Gradient accumulation steps = {args.gradient_accumulation_steps}\\n"\\
f"{chr(92)} / Data Parallel GPUs = {args.world_size} | Total batch size = {total_train_batch_size:,}\\n"\\
f' "-____-" Number of trainable parameters = {get_model_param_count(model, trainable_only=True):,}/{get_model_param_count(model)}'
logger.warning(debug_info)
import subprocess, re, gc
for _ in range(3):