HunYuanDiT switch to local tokenizer

HunYuanDiT/mt5_tokenizer/config.json

@@ -0,0 +1,33 @@
+{
+  "_name_or_path": "mt5",
+  "architectures": [
+    "MT5ForConditionalGeneration"
+  ],
+  "classifier_dropout": 0.0,
+  "d_ff": 5120,
+  "d_kv": 64,
+  "d_model": 2048,
+  "decoder_start_token_id": 0,
+  "dense_act_fn": "gelu_new",
+  "dropout_rate": 0.1,
+  "eos_token_id": 1,
+  "feed_forward_proj": "gated-gelu",
+  "initializer_factor": 1.0,
+  "is_encoder_decoder": true,
+  "is_gated_act": true,
+  "layer_norm_epsilon": 1e-06,
+  "model_type": "mt5",
+  "num_decoder_layers": 24,
+  "num_heads": 32,
+  "num_layers": 24,
+  "output_past": true,
+  "pad_token_id": 0,
+  "relative_attention_max_distance": 128,
+  "relative_attention_num_buckets": 32,
+  "tie_word_embeddings": false,
+  "tokenizer_class": "T5Tokenizer",
+  "torch_dtype": "float16",
+  "transformers_version": "4.40.2",
+  "use_cache": true,
+  "vocab_size": 250112
+}

HunYuanDiT/mt5_tokenizer/special_tokens_map.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}

HunYuanDiT/mt5_tokenizer/tokenizer_config.json

@@ -0,0 +1 @@
+{"eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>", "extra_ids": 0, "additional_special_tokens": null, "special_tokens_map_file": "/home/patrick/.cache/torch/transformers/685ac0ca8568ec593a48b61b0a3c272beee9bc194a3c7241d15dcadb5f875e53.f76030f3ec1b96a8199b2593390c610e76ca8028ef3d24680000619ffb646276", "tokenizer_file": null, "name_or_path": "google/mt5-small"}

HunYuanDiT/tenc.py

@@ -103,17 +103,26 @@ class EXM_HyDiT_Tenc_Temp:
 				device         = self.load_device,
 				dtype          = self.dtype,
 			)
-			tokenizer_args = {"subfolder": "t2i/mt5"}
+			tokenizer_args = {"subfolder": "t2i/mt5"} # web
+			tokenizer_path = os.path.join( # local
+				os.path.dirname(os.path.realpath(__file__)),
+				"mt5_tokenizer",
+			)
 		else:
 			self.cond_stage_model = hyCLIPModel(
 				device         = self.load_device,
 				dtype          = self.dtype,
 			)
-			tokenizer_args = {"subfolder": "t2i/tokenizer",}
+			tokenizer_args = {"subfolder": "t2i/tokenizer",} # web
-		self.tokenizer = AutoTokenizer.from_pretrained(
+			tokenizer_path = os.path.join( # local
-			"Tencent-Hunyuan/HunyuanDiT",
+				os.path.dirname(os.path.realpath(__file__)),
-			**tokenizer_args
+				"tokenizer",
 			)
+		# self.tokenizer = AutoTokenizer.from_pretrained(
+			# "Tencent-Hunyuan/HunyuanDiT",
+			# **tokenizer_args
+		# )
+		self.tokenizer = AutoTokenizer.from_pretrained(tokenizer_path)
 		self.patcher = comfy.model_patcher.ModelPatcher(
 			self.cond_stage_model,
 			load_device    = self.load_device,

HunYuanDiT/tokenizer/config.json

@@ -0,0 +1,34 @@
+{
+  "_name_or_path": "hfl/chinese-roberta-wwm-ext-large",
+  "architectures": [
+    "BertModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "directionality": "bidi",
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "output_past": true,
+  "pad_token_id": 0,
+  "pooler_fc_size": 768,
+  "pooler_num_attention_heads": 12,
+  "pooler_num_fc_layers": 3,
+  "pooler_size_per_head": 128,
+  "pooler_type": "first_token_transform",
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.22.1",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 47020
+}

HunYuanDiT/tokenizer/special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

HunYuanDiT/tokenizer/tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "name_or_path": "hfl/chinese-roberta-wwm-ext",
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "special_tokens_map_file": "/home/chenweifeng/.cache/huggingface/hub/models--hfl--chinese-roberta-wwm-ext/snapshots/5c58d0b8ec1d9014354d691c538661bf00bfdb44/special_tokens_map.json",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]",
+  "model_max_length": 77
+}