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

narrow Nemotron trust_remote_code to nemotron_h/nemotron-3-nano, add to export worker

Browse source
This commit is contained in:
Roland Tannous 2026-04-06 22:45:22 +00:00
commit 721bdadafa
3 changed files with 21 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

13
studio/backend/core/export/worker.py
View file

@ -206,6 +206,19 @@ def _handle_load(backend, cmd: dict, resp_queue: Any) -> None:
load_in_4bit = cmd.get("load_in_4bit", True)
trust_remote_code = cmd.get("trust_remote_code", False)
# Auto-enable trust_remote_code for NemotronH/Nano models.
if not trust_remote_code:
_NEMOTRON_TRUST_SUBSTRINGS = ("nemotron_h", "nemotron-h", "nemotron-3-nano")
_cp_lower = checkpoint_path.lower()
if any(sub in _cp_lower for sub in _NEMOTRON_TRUST_SUBSTRINGS) and (
_cp_lower.startswith("unsloth/") or _cp_lower.startswith("nvidia/")
):
trust_remote_code = True
logger.info(
"Auto-enabled trust_remote_code for Nemotron model: %s",
checkpoint_path,
)
try:
_send_response(
resp_queue,

6
studio/backend/core/inference/worker.py
View file

@ -287,14 +287,16 @@ def _handle_load(backend, config: dict, resp_queue: Any) -> None:
except Exception as e:
logger.warning("Could not read adapter_config.json: %s", e)
# Auto-enable trust_remote_code for Nemotron models only.
# Auto-enable trust_remote_code for NemotronH/Nano models only.
# NemotronH has config parsing bugs requiring trust_remote_code=True.
# Other transformers 5.x models are native and do NOT need it.
# NOTE: Must NOT match Llama-Nemotron (standard Llama architecture).
_NEMOTRON_TRUST_SUBSTRINGS = ("nemotron_h", "nemotron-h", "nemotron-3-nano")
trust_remote_code = config.get("trust_remote_code", False)
if not trust_remote_code:
model_name = config["model_name"]
_mn_lower = model_name.lower()
if "nemotron" in _mn_lower and (
if any(sub in _mn_lower for sub in _NEMOTRON_TRUST_SUBSTRINGS) and (
_mn_lower.startswith("unsloth/") or _mn_lower.startswith("nvidia/")
):
trust_remote_code = True

6
studio/backend/core/training/worker.py
View file

@ -1828,14 +1828,16 @@ def run_training_process(
)
return
# ── 1a. Auto-enable trust_remote_code for Nemotron models ──
# ── 1a. Auto-enable trust_remote_code for NemotronH/Nano models ──
# NemotronH has config parsing bugs in transformers that require
# trust_remote_code=True as a workaround. Other transformers 5.x models
# (Qwen3.5, Gemma 4, etc.) are native and do NOT need it — enabling it
# bypasses the compiler (disabling fused CE).
# NOTE: Must NOT match Llama-Nemotron (standard Llama architecture).
_NEMOTRON_TRUST_SUBSTRINGS = ("nemotron_h", "nemotron-h", "nemotron-3-nano")
_lowered = model_name.lower()
if (
"nemotron" in _lowered
any(sub in _lowered for sub in _NEMOTRON_TRUST_SUBSTRINGS)
and (_lowered.startswith("unsloth/") or _lowered.startswith("nvidia/"))
and not config.get("trust_remote_code", False)
):