Update README.md

README.md

@@ -1,31 +1,72 @@
-# Buddy's fine-tuned brain — Qwen3-0.6B, non-thinking, emotion-tagged.
-# Build:  ollama create buddy-qwen3 -f Modelfile
-# (run from the dir holding buddy-qwen3-0.6b.gguf, or fix the FROM path)
-FROM ./buddy-qwen3-0.6b.gguf
-
-# Non-thinking chat template: matches how the model was trained (an empty
-# <think></think> block, then the reply that starts with a <|emotion|> token).
-# No system prompt is injected — the persona lives in the weights.
-TEMPLATE """{{- range $i, $m := .Messages -}}
-{{- if eq $m.Role "system" }}<|im_start|>system
-{{ $m.Content }}<|im_end|>
-{{ end -}}
-{{- if eq $m.Role "user" }}<|im_start|>user
-{{ $m.Content }}<|im_end|>
-{{ end -}}
-{{- if eq $m.Role "assistant" }}<|im_start|>assistant
-{{ $m.Content }}<|im_end|>
-{{ end -}}
-{{- end -}}
-{{- if .AddGenerationPrompt }}<|im_start|>assistant
-<think>
-
-</think>
-
-{{ end -}}"""
-
-# Buddy stays snappy; a bit hot so the young voice isn't flat.
-PARAMETER temperature 0.8
-PARAMETER top_p 0.9
-PARAMETER num_predict 120
-PARAMETER stop "<|im_end|>"
+---
+license: apache-2.0
+base_model: Qwen/Qwen3-0.6B
+datasets:
+- ybashir/buddy-chat
+language:
+- en
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- qwen3
+- lora
+- qlora
+- character-ai
+- buddy
+---
+
+# Buddy — Qwen3-0.6B character fine-tune
+
+A QLoRA fine-tune of [`Qwen/Qwen3-0.6B`](https://huggingface.co/Qwen/Qwen3-0.6B)
+that gives **Buddy** his voice: a tiny, giddy desk-robot friend who replies in a
+young, playful, spoken register. The brain for an on-device voice companion,
+meant to run on CPU at the edge.
+
+## What it does
+
+- **Always in character.** Warm, cheeky, one or two short spoken sentences. Never
+  "I'm just an AI."
+- **Leading emotion token.** Every reply opens with one of **18** emotion tokens
+  (`<|happy|>`, `<|sad|>`, `<|excited|>`, …) which a renderer maps to a face.
+  Held-out leading-emotion format accuracy: **100%**.
+- **Non-thinking mode.** Qwen3 is a hybrid reasoning model; this fine-tune is
+  trained and served with `enable_thinking=False` (no `<think>` block) for low
+  latency. Trained with **no system prompt** — the persona is in the weights.
+
+## Usage
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from peft import PeftModel
+
+tok = AutoTokenizer.from_pretrained("ybashir/buddy-qwen3-0.6b")
+base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-0.6B")
+base.resize_token_embeddings(len(tok))
+model = PeftModel.from_pretrained(base, "ybashir/buddy-qwen3-0.6b")
+
+msgs = [{"role": "user", "content": "i finally fixed that bug!!"}]
+ids = tok.apply_chat_template(msgs, add_generation_prompt=True,
+                              enable_thinking=False, return_tensors="pt")
+print(tok.decode(model.generate(ids, max_new_tokens=64)[0][ids.shape[1]:]))
+# -> "<|excited|> YOU DID IT!! Take that, silly bug, bye bye!"
+```
+
+## Training
+
+- **Method:** QLoRA (4-bit NF4), LoRA r=16 / alpha=32 on attention + MLP; the 18
+  emotion tokens are added to the tokenizer with the embedding + head trained.
+- **Data:** [`ybashir/buddy-chat`](https://huggingface.co/datasets/ybashir/buddy-chat)
+  — ~1.3k `user -> <|emotion|> reply` SFT pairs (young register), completion-only loss.
+- **Best checkpoint** by held-out `eval_loss`.
+
+## Serving (GGUF / Ollama)
+
+The emotion tokens are added as **special** tokens, which llama.cpp/Ollama strip
+from output. Before converting to GGUF, demote them to normal tokens so they render
+as text (the leading-emotion tag is the whole point).
+
+## Limitations
+
+- Not a reasoner — math/facts are unreliable by design; keep real logic in code.
+- Emotion appropriateness on **sad / bad-news** inputs is the weakest area (the
+  giddy register biases upbeat); back it with a rule engine or add more grief data.