update

Author: DN6

src/diffusers/pipelines/hidream_image/pipeline_hidream_image.py

@@ -36,11 +36,11 @@
     Examples:
         ```py
         >>> import torch
-        >>> from transformers import PreTrainedTokenizerFast, LlamaForCausalLM
-        >>> from diffusers import UniPCMultistepScheduler, HiDreamImagePipeline
+        >>> from transformers import AutoTokenizer, LlamaForCausalLM
+        >>> from diffusers import HiDreamImagePipeline
 
 
-        >>> tokenizer_4 = PreTrainedTokenizerFast.from_pretrained("meta-llama/Meta-Llama-3.1-8B-Instruct")
+        >>> tokenizer_4 = AutoTokenizer.from_pretrained("meta-llama/Meta-Llama-3.1-8B-Instruct")
         >>> text_encoder_4 = LlamaForCausalLM.from_pretrained(
         ...     "meta-llama/Meta-Llama-3.1-8B-Instruct",
         ...     output_hidden_states=True,
@@ -901,6 +901,7 @@ def __call__(
             pooled_prompt_embeds=pooled_prompt_embeds,
             negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
             device=device,
+            dtype=self.dtype,
             num_images_per_prompt=num_images_per_prompt,
             max_sequence_length=max_sequence_length,
             lora_scale=lora_scale,

tests/quantization/gguf/test_gguf.py

@@ -557,3 +557,23 @@ class HiDreamGGUFSingleFileTests(GGUFSingleFileTesterMixin, unittest.TestCase):
     torch_dtype = torch.bfloat16
     model_cls = HiDreamImageTransformer2DModel
     expected_memory_use_in_gb = 8
+
+    def get_dummy_inputs(self):
+        return {
+            "hidden_states": torch.randn((1, 16, 128, 128), generator=torch.Generator("cpu").manual_seed(0)).to(
+                torch_device, self.torch_dtype
+            ),
+            "encoder_hidden_states_t5": torch.randn(
+                (1, 128, 4096),
+                generator=torch.Generator("cpu").manual_seed(0),
+            ).to(torch_device, self.torch_dtype),
+            "encoder_hidden_states_llama3": torch.randn(
+                (32, 1, 128, 4096),
+                generator=torch.Generator("cpu").manual_seed(0),
+            ).to(torch_device, self.torch_dtype),
+            "pooled_embeds": torch.randn(
+                (1, 2048),
+                generator=torch.Generator("cpu").manual_seed(0),
+            ).to(torch_device, self.torch_dtype),
+            "timestep": torch.tensor([1]).to(torch_device, self.torch_dtype),
+        }