Upload 大模型.txt

大模型.txt

@@ -0,0 +1,30 @@
+
+prompt+Image index + question
+
+
+tensor([[    1,   319, 13563,  1546,   263, 12758,  5199,   322,   385, 23116,
+         21082, 20255, 29889,   450, 20255,  4076,  8444, 29892, 13173, 29892,
+           322,  1248,   568,  6089,   304,   278,  5199, 29915, 29879,  5155,
+         29889,  3148,  1001, 29901, 29871,  -200, 29871,    13,  5618,  1203,
+          4318,  2444,  1556, 22910, 29973,   319, 29901,  9795,   342,   374,
+           550,   350, 29901,   824,   342,   432,  2873,   315, 29901,  3018,
+          2416,  3578,   360, 29901, 29352, 29889,   319,  1799,  9047, 13566,
+         29901]], device='cuda:0')
+
+<s> A chat between a curious human and an artificial intelligence assistant
+646
+
+576
+
+input 71  prompt+Image index token（1） + question
+
+input_embedding （1,646,4096）
+num image token 646 - （71 - 1） = 576
+
+
+squence  176  
+score  175   shape (1, 32000)
+attentions 175  32   shape (1,32,1,646+num_generate_token)
+past kv 32 (1,32(head),820(squence length),128)
+
+生成时output.sequences比hidden_state长度多1，因为最后会拼接一个<\s>即end_of_sequence的特殊token以表示序列结束。且模型输入后会先生成一个<s>即start_of_sequence，