Closed brothermaster closed 3 months ago
All input embeddings were trained in the "initialize_graph_tokenizer" method at Self-Supervised Instruction Tuning stage , but only "-num_new_token" were loaded in Task-Specific Instruction Tuning stage while embedding of other tokens are not used. Question 1: Why don't you use all embeddings that were trained? Question 2: Actually, embeddings of "num_new_token" of graph are not input into LLM.Text embeddings and aligned graph representation are concatenated and input into LLM. Why do you train graph tokens?
您好,我想请问一下,您知道Graph token是怎么得到的么
All input embeddings were trained in the "initialize_graph_tokenizer" method at Self-Supervised Instruction Tuning stage , but only "-num_new_token" were loaded in Task-Specific Instruction Tuning stage while embedding of other tokens are not used. Question 1: Why don't you use all embeddings that were trained? Question 2: Actually, embeddings of "num_new_token" of graph are not input into LLM.Text embeddings and aligned graph representation are concatenated and input into LLM. Why do you train graph tokens?
您好,我想请问一下,您知道Graph token是怎么得到的么
All input embeddings were trained in the "initialize_graph_tokenizer" method at Self-Supervised Instruction Tuning stage , but only "-num_new_token" were loaded in Task-Specific Instruction Tuning stage while embedding of other tokens are not used. Question 1: Why don't you use all embeddings that were trained? Question 2: Actually, embeddings of "num_new_token" of graph are not input into LLM.Text embeddings and aligned graph representation are concatenated and input into LLM. Why do you train graph tokens?
您好,我想请问一下,您知道Graph token是怎么得到的么
在这个位置:
好诶,感谢!
All input embeddings were trained in the "initialize_graph_tokenizer" method at Self-Supervised Instruction Tuning stage , but only "-num_new_token" were loaded in Task-Specific Instruction Tuning stage while embedding of other tokens are not used. Question 1: Why don't you use all embeddings that were trained?
Question 2: Actually, embeddings of "num_new_token" of graph are not input into LLM.Text embeddings and aligned graph representation are concatenated and input into LLM. Why do you train graph tokens?