增加llama3.2

Llama-3.2-1B-Instruct/bert4torch_config.json

@@ -0,0 +1,29 @@
+{
+	"model": "llama",
+	"template": "llama3",
+	"hidden_size": 2048,
+    "intermediate_size": 8192, 
+	"num_attention_heads": 32,
+	"num_hidden_layers": 16,
+	"num_key_value_heads": 8,
+	"hidden_act": "silu",
+	"vocab_size": 128256,
+	"segment_vocab_size": 0,
+	"skip_init": true,
+	"layer_norm_eps": 1e-05,
+	"rope_rank": "updown",
+	"rope_theta": 500000.0,
+	"rope_scaling": {
+	  "factor": 32.0,
+      "low_freq_factor": 1.0,
+      "high_freq_factor": 4.0,
+      "original_max_position_embeddings": 8192,
+      "rope_type": "llama3"
+    },
+	"max_position_embeddings": 131072,
+	"torch_dtype": "bfloat16",
+	"tie_word_embeddings": true,
+	"attention_probs_dropout_prob": 0.0,
+    "convert_lm_logits_dtype": "float32",
+	"generation_config": {"tokenizer_config":  {"skip_special_tokens": true, "add_special_tokens": false}, "eos_token_id": [128001, 128008, 128009]}
+}

Llama-3.2-1B/bert4torch_config.json

@@ -0,0 +1,29 @@
+{
+	"model": "llama",
+	"template": "llama3",
+	"hidden_size": 2048,
+    "intermediate_size": 8192, 
+	"num_attention_heads": 32,
+	"num_hidden_layers": 16,
+	"num_key_value_heads": 8,
+	"hidden_act": "silu",
+	"vocab_size": 128256,
+	"segment_vocab_size": 0,
+	"skip_init": true,
+	"layer_norm_eps": 1e-05,
+	"rope_rank": "updown",
+	"rope_theta": 500000.0,
+	"rope_scaling": {
+	  "factor": 32.0,
+      "low_freq_factor": 1.0,
+      "high_freq_factor": 4.0,
+      "original_max_position_embeddings": 8192,
+      "rope_type": "llama3"
+    },
+	"max_position_embeddings": 131072,
+	"torch_dtype": "bfloat16",
+	"tie_word_embeddings": true,
+	"attention_probs_dropout_prob": 0.0,
+    "convert_lm_logits_dtype": "float32",
+	"generation_config": {"tokenizer_config":  {"skip_special_tokens": true, "add_special_tokens": false}, "eos_token_id": [128001, 128008, 128009]}
+}

Llama-3.2-3B-Instruct/bert4torch_config.json

@@ -0,0 +1,29 @@
+{
+	"model": "llama",
+	"template": "llama3",
+	"hidden_size": 3072,
+    "intermediate_size": 8192, 
+	"num_attention_heads": 24,
+	"num_hidden_layers": 28,
+	"num_key_value_heads": 8,
+	"hidden_act": "silu",
+	"vocab_size": 128256,
+	"segment_vocab_size": 0,
+	"skip_init": true,
+	"layer_norm_eps": 1e-05,
+	"rope_rank": "updown",
+	"rope_theta": 500000.0,
+	"rope_scaling": {
+	  "factor": 32.0,
+      "low_freq_factor": 1.0,
+      "high_freq_factor": 4.0,
+      "original_max_position_embeddings": 8192,
+      "rope_type": "llama3"
+    },
+	"max_position_embeddings": 131072,
+	"torch_dtype": "bfloat16",
+	"tie_word_embeddings": true,
+	"attention_probs_dropout_prob": 0.0,
+    "convert_lm_logits_dtype": "float32",
+	"generation_config": {"tokenizer_config":  {"skip_special_tokens": true, "add_special_tokens": false}, "eos_token_id": [128001, 128008, 128009]}
+}

Llama-3.2-3B/bert4torch_config.json

@@ -0,0 +1,29 @@
+{
+	"model": "llama",
+	"template": "llama3",
+	"hidden_size": 3072,
+    "intermediate_size": 8192, 
+	"num_attention_heads": 24,
+	"num_hidden_layers": 28,
+	"num_key_value_heads": 8,
+	"hidden_act": "silu",
+	"vocab_size": 128256,
+	"segment_vocab_size": 0,
+	"skip_init": true,
+	"layer_norm_eps": 1e-05,
+	"rope_rank": "updown",
+	"rope_theta": 500000.0,
+	"rope_scaling": {
+	  "factor": 32.0,
+      "low_freq_factor": 1.0,
+      "high_freq_factor": 4.0,
+      "original_max_position_embeddings": 8192,
+      "rope_type": "llama3"
+    },
+	"max_position_embeddings": 131072,
+	"torch_dtype": "bfloat16",
+	"tie_word_embeddings": true,
+	"attention_probs_dropout_prob": 0.0,
+    "convert_lm_logits_dtype": "float32",
+	"generation_config": {"tokenizer_config":  {"skip_special_tokens": true, "add_special_tokens": false}, "eos_token_id": [128001, 128008, 128009]}
+}