wake up, eric

src/cartman.py

@@ -3,31 +3,34 @@ import json
 
 url = 'https://doordesk.net/chat'
 
-def cartman_speak(user_message):
+def cartman_respond(user_message):
     message = {'Message': user_message}
     response = requests.post(url,json.dumps(message))
     return response.json().get('Cartman')
 
+from transformers.models.auto.modeling_auto import AutoModelForCausalLM
+from transformers.models.auto.tokenization_auto import AutoTokenizer
 
-# from transformers.models.auto.tokenization_auto import AutoTokenizer
-# from transformers.models.auto.modeling_auto import AutoModelForCausalLM
-# import torch
-# 
-# tokenizer = AutoTokenizer.from_pretrained('microsoft/DialoGPT-large')
-# model = AutoModelForCausalLM.from_pretrained('../southpark/output-medium')
-# 
-# def cartman_speak(user_message):
-#     new_user_input_ids = tokenizer.encode(user_message + tokenizer.eos_token, return_tensors='pt')
-#     bot_output = new_user_input_ids
-#     bot_input_ids = torch.cat([new_user_input_ids, bot_output])
-#     bot_output = model.generate(
-#         bot_input_ids, max_length= 200,
-#         pad_token_id=tokenizer.eos_token_id,
-#         no_repeat_ngram_size=3,
-#         do_sample=True,
-#         top_k=100,
-#         top_p=0.7,
-#         temperature=.8
-#     )
-# 
-#     return '{}'.format(tokenizer.decode(bot_output[:,bot_input_ids.shape[-1]:][0], skip_special_tokens=True))
+tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-large")
+model = AutoModelForCausalLM.from_pretrained("../chatbots/southpark/cartman/models/output-medium")
+
+def cartman_speak(input_text):
+    input_ids = tokenizer(input_text + tokenizer.eos_token, return_tensors="pt").input_ids
+    outputs = model.generate(
+            input_ids,
+            pad_token_id=tokenizer.eos_token_id,
+            max_new_tokens = 200,
+            num_beams = 8,
+            num_beam_groups = 4,
+            no_repeat_ngram_size=3,
+            length_penalty = 1.4,
+            diversity_penalty = 0,
+            repetition_penalty = 2.1,
+            early_stopping = True,
+
+          # do_sample = True,
+          # top_k = 100,
+          # top_p = 0.7,
+          # temperature = 0.8,
+            )
+    return tokenizer.decode(outputs[:, input_ids.shape[-1]:][0], skip_special_tokens=True)

test/flantest.py

@@ -4,7 +4,7 @@ from transformers.models.t5.modeling_t5 import T5ForConditionalGeneration
 
 device = torch.device("cuda")
 tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-xl")
-model = T5ForConditionalGeneration.from_pretrained("google/flan-t5-xxl").cuda()
+model = T5ForConditionalGeneration.from_pretrained("google/flan-t5-xl").cuda()
 
 run = True
 while run:
@@ -12,7 +12,7 @@ while run:
     if input_text in 'q':
         run = False
         break
-    input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to("cuda")
+    input_ids = tokenizer.encode(input_text, return_tensors="pt").to("cuda")
 
     outputs = model.generate(input_ids)
-    print(tokenizer.decode(outputs[0]))
+    print(tokenizer.batch_decode(outputs, skip_special_tokens=True)[0])

test/gpt-jt_test.py

@@ -0,0 +1,18 @@
+
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+device = torch.device("cuda")
+tokenizer = AutoTokenizer.from_pretrained("togethercomputer/GPT-JT-6B-v1")
+model = AutoModelForCausalLM.from_pretrained("togethercomputer/GPT-JT-6B-v1").cuda()
+
+run = True
+while run:
+    input_text = input('>> ')
+    if input_text in 'q':
+        run = False
+        break
+    input_ids = tokenizer.encode(input_text, return_tensors="pt").to("cuda")
+
+    outputs = model.generate(input_ids)
+    print(tokenizer.batch_decode(outputs, skip_special_tokens=True)[0])

test/gptjtest.py

@@ -0,0 +1,17 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+device = torch.device("cuda")
+tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B")
+model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-j-6B").cuda()
+
+run = True
+while run:
+    input_text = input('>> ')
+    if input_text in 'q':
+        run = False
+        break
+    input_ids = tokenizer.encode(input_text, return_tensors="pt").to("cuda")
+
+    outputs = model.generate(input_ids)
+    print(tokenizer.batch_decode(outputs, skip_special_tokens=True)[0])