Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Make Qwen 8B work with TextGenerate node. (#13160)

comfyanonymous2a1f4026018f
diff --git a/comfy/ops.py b/comfy/ops.py
index ca25693db..b5cd1d47e 100644
--- a/comfy/ops.py
+++ b/comfy/ops.py
@@ -928,6 +928,7 @@ def mixed_precision_ops(quant_config={}, compute_dtype=torch.bfloat16, full_prec
weight = state_dict.pop(weight_key, None)
if weight is None:
logging.warning(f"Missing weight for layer {layer_name}")
+ self.weight = None
return
manually_loaded_keys = [weight_key]
@@ -1034,6 +1035,9 @@ def mixed_precision_ops(quant_config={}, compute_dtype=torch.bfloat16, full_prec
if self.bias is not None:
sd["{}bias".format(prefix)] = self.bias
+ if self.weight is None:
+ return sd
+
if isinstance(self.weight, QuantizedTensor):
sd_out = self.weight.state_dict("{}weight".format(prefix))
for k in sd_out:
diff --git a/comfy/text_encoders/llama.py b/comfy/text_encoders/llama.py
index d93926648..06f2fbf74 100644
--- a/comfy/text_encoders/llama.py
+++ b/comfy/text_encoders/llama.py
@@ -224,7 +224,7 @@ class Qwen3_8BConfig:
k_norm = "gemma3"
rope_scale = None
final_norm: bool = True
- lm_head: bool = False
+ lm_head: bool = True
stop_tokens = [151643, 151645]
@dataclass
@@ -912,6 +912,9 @@ class BaseGenerate:
class BaseQwen3:
def logits(self, x):
input = x[:, -1:]
+ if self.model.config.lm_head:
+ return self.model.lm_head(input)
+
module = self.model.embed_tokens
offload_stream = None