From bae6bff5cec69460680673a206b35d6f9dfa763a Mon Sep 17 00:00:00 2001
From: Dani <dsapelli@yahoo.com>
Date: Sun, 27 Apr 2025 14:03:37 -0400
Subject: [PATCH] Fixing how she replies.

---
 model/brain.py     | 42 +++++++++++++++++-------------------------
 model/tokenizer.py |  2 ++
 2 files changed, 19 insertions(+), 25 deletions(-)

diff --git a/model/brain.py b/model/brain.py
index 2f265ec..a6c3b48 100644
--- a/model/brain.py
+++ b/model/brain.py
@@ -13,40 +13,32 @@ recent_dreams = []
 @torch.no_grad()
 def generate_response():
     model.eval()
-    context_texts = get_recent_context(10)
-    seed_text = " ".join(context_texts[-1:])
-    tokens = tokenizer.tokenize(seed_text)
-
-    input_tensor = torch.tensor(tokens, dtype=torch.long, device=DEVICE).unsqueeze(0)
-
+    seed = torch.randint(0, model.head.out_features, (1, 1), device=DEVICE)
+    input_ids = seed
     output_tokens = []
-    max_tokens = 32
 
-    for _ in range(max_tokens):
-        output = model(input_tensor)
-        logits = output[:, -1, :].squeeze(0)
+    for _ in range(50):  # Max 50 tokens (short sentences)
+        output = model(input_ids)
+        next_token_logits = output[:, -1, :] / 0.8  # temperature 0.8
 
-        # Apply temperature (soft randomness)
-        temperature = 0.8
-        logits = logits / temperature
+        # Top-K Sampling
+        top_k = 40
+        values, indices = torch.topk(next_token_logits, k=top_k)
+        probs = F.softmax(values, dim=-1)
+        sampled_idx = torch.multinomial(probs, num_samples=1)
 
-        # Top-k sampling
-        k = 10
-        topk_logits, topk_indices = torch.topk(logits, k)
-        probs = torch.nn.functional.softmax(topk_logits, dim=-1)
-        next_token = topk_indices[torch.multinomial(probs, 1)].item()
+        next_token = indices.gather(-1, sampled_idx)
 
-        output_tokens.append(next_token)
+        output_tokens.append(next_token.item())
 
-        input_tensor = torch.cat([input_tensor, torch.tensor([[next_token]], device=DEVICE)], dim=1)
+        input_ids = torch.cat([input_ids, next_token.view(1, 1)], dim=1)
 
-        # Optional: stop if next_token maps to period, question mark, or exclamation
-        next_char = tokenizer.detokenize([next_token])
-        if any(p in next_char for p in [".", "?", "!"]):
+        # Break if punctuation (end of sentence)
+        word = tokenizer.detokenize(next_token.item())
+        if word in [".", "!", "?"]:
             break
 
-    text = tokenizer.detokenize(output_tokens)
-    return text
+    return tokenizer.detokenize(output_tokens)
 
 
 def score_sentence(sentence: str) -> float:
diff --git a/model/tokenizer.py b/model/tokenizer.py
index 5983b3f..663bbb2 100644
--- a/model/tokenizer.py
+++ b/model/tokenizer.py
@@ -36,4 +36,6 @@ class Tokenizer:
         return tokens
 
     def detokenize(self, tokens):
+        if isinstance(tokens, int):
+            tokens = [tokens]
         return " ".join(self.reverse_vocab.get(t, "<unk>") for t in tokens)