From a345a2acd2df85910669a84a81926296b632a36c Mon Sep 17 00:00:00 2001
From: oobabooga <112222186+oobabooga@users.noreply.github.com>
Date: Fri, 3 Mar 2023 15:16:55 -0300
Subject: [PATCH] Add a tokenizer placeholder

---
 modules/text_generation.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/modules/text_generation.py b/modules/text_generation.py
index 6537512..bd018c0 100644
--- a/modules/text_generation.py
+++ b/modules/text_generation.py
@@ -21,8 +21,11 @@ def get_max_prompt_length(tokens):
     return max_length
 
 def encode(prompt, tokens_to_generate=0, add_special_tokens=True):
+
+    # These models do not have explicit tokenizers for now, so
+    # we return an estimate on the number of tokens
     if shared.is_RWKV or shared.is_LLaMA:
-        return prompt
+        return np.zeros((1, len(prompt)//5))
 
     input_ids = shared.tokenizer.encode(str(prompt), return_tensors='pt', truncation=True, max_length=get_max_prompt_length(tokens_to_generate), add_special_tokens=add_special_tokens)
     if shared.args.cpu: