search

yolain / ComfyUI-Easy-Use

In order to make it easier to use the ComfyUI, I have made some optimizations and integrations to some commonly used nodes.
GNU General Public License v3.0
1.04k stars 71 forks source link

Add support for T5 ENCODER GGUF #336

Open enternalsaga opened 2 months ago

enternalsaga commented 2 months ago

hi, since native comfyui already supported t5 GUFF text encoder to speed up loading Flux model link can you also add support for it? the link is here: https://huggingface.co/city96/t5-v1_1-xxl-encoder-gguf/tree/main When i tried dual clip GUFF loader using fluxloader there is error: image

`Error occurred when executing easy fluxLoader:

module 'gguf.quants' has no attribute 'dequantize'

File "I:\ComfyUI\execution.py", line 316, in execute output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "I:\ComfyUI\execution.py", line 191, in get_output_data return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb) File "I:\ComfyUI\execution.py", line 168, in _map_node_over_list process_inputs(input_dict, i) File "I:\ComfyUI\execution.py", line 157, in process_inputs results.append(getattr(obj, func)(inputs)) File "I:\ComfyUI\custom_nodes\15_ComfyUI-Easy-Use\py\easyNodes.py", line 1977, in fluxloader return super().adv_pipeloader(ckpt_name, 'Default', vae_name, 0, File "I:\ComfyUI\custom_nodes\15_ComfyUI-Easy-Use\py\easyNodes.py", line 937, in adv_pipeloader positive_embeddings_final, positive_wildcard_prompt, model, clip = prompt_to_cond('positive', model, clip, clip_skip, lora_stack, positive, positive_token_normalization, positive_weight_interpretation, a1111_prompt_style, my_unique_id, prompt, easyCache, model_type=model_type) File "I:\ComfyUI\custom_nodes\15_ComfyUI-Easy-Use\py\libs\conditioning.py", line 17, in prompt_to_cond embeddings_final, = CLIPTextEncode().encode(clip, text) File "I:\ComfyUI\nodes.py", line 65, in encode output = clip.encode_from_tokens(tokens, return_pooled=True, return_dict=True) File "I:\ComfyUI\comfy\sd.py", line 126, in encode_from_tokens o = self.cond_stage_model.encode_token_weights(tokens) File "I:\ComfyUI\comfy\sdxl_clip.py", line 58, in encode_token_weights g_out, g_pooled = self.clip_g.encode_token_weights(token_weight_pairs_g) File "I:\ComfyUI\comfy\sd1_clip.py", line 41, in encode_token_weights o = self.encode(to_encode) File "I:\ComfyUI\comfy\sd1_clip.py", line 229, in encode return self(tokens) File "I:\ComfyUI\venv\lib\site-packages\torch\nn\modules\module.py", line 1716, in _wrapped_call_impl return self._call_impl(*args, *kwargs) File "I:\ComfyUI\venv\lib\site-packages\torch\nn\modules\module.py", line 1727, in _call_impl return forward_call(args, kwargs) File "I:\ComfyUI\comfy\sd1_clip.py", line 201, in forward outputs = self.transformer(tokens, attention_mask_model, intermediate_output=self.layer_idx, final_layer_norm_intermediate=self.layer_norm_hidden_state, dtype=torch.float32) File "I:\ComfyUI\venv\lib\site-packages\torch\nn\modules\module.py", line 1716, in _wrapped_call_impl return self._call_impl(*args, kwargs) File "I:\ComfyUI\venv\lib\site-packages\torch\nn\modules\module.py", line 1727, in _call_impl return forward_call(*args, *kwargs) File "I:\ComfyUI\comfy\clip_model.py", line 136, in forward x = self.text_model(args, kwargs) File "I:\ComfyUI\venv\lib\site-packages\torch\nn\modules\module.py", line 1716, in _wrapped_call_impl return self._call_impl(*args, *kwargs) File "I:\ComfyUI\venv\lib\site-packages\torch\nn\modules\module.py", line 1727, in _call_impl return forward_call(args, **kwargs) File "I:\ComfyUI\comfy\clip_model.py", line 112, in forward x, i = self.encoder(x, mask=mask, intermediate_output=intermediate_output) File "I:\ComfyUI\venv\lib\site-packages\torch\nn\modules\module.py", line 1716, in _wrapped_call_impl return self.`

EnragedAntelope commented 1 month ago

Hi, I am also having issues with Flux and XY Grid. But I want to note in your image, you have left the Clip loader setting at SDXL instead of choosing Flux. I do this all the time :) but it definitely can break workflows so I thought I should let you know.