Skip to content

enable studio for intel GPU #13

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
danielhanchen wants to merge 15 commits into from
Closed

enable studio for intel GPU #13

Show file tree
Hide file tree
Changes from 11 commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
9db527a
add intel GPU for unsloth studio
leizhenyuan Mar 31, 2026
f1c426b
clean stash
leizhenyuan Mar 31, 2026
dc55c95
remove unuse code
leizhenyuan Mar 31, 2026
5487a1b
remove rl changes
leizhenyuan Mar 31, 2026
72eced4
remove unuse code
leizhenyuan Mar 31, 2026
f679dca
remove unuse code
leizhenyuan Mar 31, 2026
8dd4290
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Mar 31, 2026
03e15c7
Fix xpu-smi metric parsing, dead code, and type inconsistency
danielhanchen Apr 3, 2026
3f8dc0d
Add shutil.which guard before xpu-smi subprocess call
danielhanchen Apr 3, 2026
568a1a4
Improve get_visible_gpu_count() ZE_AFFINITY_MASK handling
danielhanchen Apr 3, 2026
9f9637e
Fix xpu-smi metrics, device resolution, and XPU OOM detection
danielhanchen Apr 8, 2026
393a70d
Wire XPU through gpu-id pinning and visibility, restore CPU OOM detec…
danielhanchen Apr 8, 2026
2cf6d02
Unblock XPU gpu_ids selection and harden OOM/autocast edge cases
danielhanchen Apr 8, 2026
5fbcb0f
Enable multi-XPU sharding and auto-select, tighten XPU edge cases
danielhanchen Apr 8, 2026
9ce735d
Align XPU wildcard mask with CUDA UUID path so multi-GPU sharding tri…
danielhanchen Apr 8, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 9 additions & 1 deletion studio/backend/core/inference/inference.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1643,7 +1643,15 @@ def _generate_dac(
+ "<|text_end|>\n<|audio_start|><|global_features_start|>\n"
)
with torch.inference_mode():
with torch.amp.autocast("cuda", dtype = model.dtype):
# Derive the autocast device from the loaded model, not from the
# global backend: a CPU-fallback DAC on an XPU/CUDA host must not
# open a GPU autocast context around CPU tensors.
device_type = (
model.device.type
if hasattr(model.device, "type")
else str(model.device).split(":", 1)[0]
)
with torch.amp.autocast(device_type, dtype = model.dtype):

@chatgpt-codex-connector chatgpt-codex-connector Bot Apr 8, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Guard CPU autocast dtype before entering DAC generation

When model.device resolves to CPU (the exact fallback this change is trying to support), this now calls torch.amp.autocast("cpu", dtype=model.dtype). CPU autocast does not accept arbitrary dtypes (notably common float32 CPU models), so entering this context can raise immediately and abort audio generation before model.generate() runs. This creates a regression for CPU-fallback DAC inference paths on non-CUDA setups; CPU should either skip autocast or use a CPU-supported autocast dtype explicitly.

Useful? React with 👍 / 👎.

inputs = tokenizer([prompt], return_tensors = "pt").to(model.device)
generated = model.generate(
**inputs,
Expand Down
13 changes: 9 additions & 4 deletions studio/backend/core/inference/llama_cpp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1438,8 +1438,9 @@ def unload_model(self) -> bool:
LlamaCppBackend._codec_mgr = None
import torch

if torch.cuda.is_available():
torch.cuda.empty_cache()
from utils.hardware import clear_gpu_cache

clear_gpu_cache()
return True

def _kill_process(self):
Expand Down Expand Up @@ -3016,7 +3017,9 @@ def init_audio_codec(self, audio_type: str) -> None:
if LlamaCppBackend._codec_mgr is None:
LlamaCppBackend._codec_mgr = AudioCodecManager()

device = "cuda" if torch.cuda.is_available() else "cpu"
from utils.hardware import get_torch_device_str

device = get_torch_device_str()
model_repo_path = None

# BiCodec needs a repo with BiCodec/ weights — download canonical SparkTTS
Expand Down Expand Up @@ -3090,7 +3093,9 @@ def generate_audio_response(

import torch

device = "cuda" if torch.cuda.is_available() else "cpu"
from utils.hardware import get_torch_device_str

device = get_torch_device_str()
return LlamaCppBackend._codec_mgr.decode(
audio_type, device, token_ids = token_ids, text = data.get("content", "")
)
24 changes: 18 additions & 6 deletions studio/backend/core/training/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1532,7 +1532,9 @@ def _preprocess_snac_dataset(self, dataset, custom_format_mapping = None):

SNAC_MODEL_NAME = "hubertsiuzdak/snac_24khz"
SNAC_SAMPLE_RATE = 24000
device = "cuda" if torch.cuda.is_available() else "cpu"
from utils.hardware import get_torch_device_str

device = get_torch_device_str()
max_length = self.max_seq_length or 2048
tokenizer = self.tokenizer

Expand Down Expand Up @@ -1708,7 +1710,9 @@ def _preprocess_snac_dataset(self, dataset, custom_format_mapping = None):
import gc

gc.collect()
torch.cuda.empty_cache()
from utils.hardware import clear_gpu_cache

clear_gpu_cache()
self._cuda_audio_used = True

if not processed_examples:
Expand Down Expand Up @@ -1736,7 +1740,9 @@ def _preprocess_bicodec_dataset(self, dataset, custom_format_mapping = None):

import subprocess

device = "cuda" if torch.cuda.is_available() else "cpu"
from utils.hardware import get_torch_device_str

device = get_torch_device_str()

# The sparktts Python package lives in the SparkAudio/Spark-TTS GitHub repo,
# NOT in the unsloth/Spark-TTS-0.5B HF model repo. Clone it if needed.
Expand Down Expand Up @@ -1936,7 +1942,9 @@ def extract_wav2vec2_features(wavs: torch.Tensor) -> torch.Tensor:
import gc

gc.collect()
torch.cuda.empty_cache()
from utils.hardware import clear_gpu_cache

clear_gpu_cache()
self._cuda_audio_used = True

if not processed_examples:
Expand Down Expand Up @@ -1971,7 +1979,9 @@ def _preprocess_dac_dataset(self, dataset, custom_format_mapping = None):
from datasets import Dataset as HFDataset
from utils.paths import ensure_dir, tmp_root

device = "cuda" if torch.cuda.is_available() else "cpu"
from utils.hardware import get_torch_device_str

device = get_torch_device_str()

# Clone OuteTTS repo (same as audio_codecs._load_dac)
import subprocess
Expand Down Expand Up @@ -2149,7 +2159,9 @@ def _preprocess_dac_dataset(self, dataset, custom_format_mapping = None):
import gc

gc.collect()
torch.cuda.empty_cache()
from utils.hardware import clear_gpu_cache

clear_gpu_cache()
self._cuda_audio_used = True

if not processed_examples:
Expand Down
2 changes: 2 additions & 0 deletions studio/backend/utils/hardware/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,7 @@
estimate_required_model_memory_gb,
auto_select_gpu_ids,
prepare_gpu_selection,
get_torch_device_str,
safe_num_proc,
safe_thread_num_proc,
dataset_map_num_proc,
Expand Down Expand Up @@ -68,6 +69,7 @@
"estimate_required_model_memory_gb",
"auto_select_gpu_ids",
"prepare_gpu_selection",
"get_torch_device_str",
"safe_num_proc",
"safe_thread_num_proc",
"dataset_map_num_proc",
Expand Down
Loading
Loading