A ~9M parameter LLM that talks like a small fish.
Python
Add --prompt for chat
Philip Whited7ee72d0e4cc
diff --git a/README.md b/README.mdindex 9ba8654..78a2c4f 100644--- a/README.md+++ b/README.md@@ -122,6 +122,14 @@ You> it is raining outsideGuppy> i think rain is the best thing about outside.```+In interactive chat mode, the conversation grows and quickly runs into the 128-token limit, reducing quality.+You can also invoke chat with a single prompt, and exit after the response:++```bash+python -m guppylm chat --prompt "tell me a joke"+```++---## Datasetdiff --git a/guppylm/__main__.py b/guppylm/__main__.pyindex 1d1f7f8..edbd42d 100644--- a/guppylm/__main__.py+++ b/guppylm/__main__.py@@ -25,20 +25,8 @@ def main():train()elif cmd == "chat":- from .inference import GuppyInference- engine = GuppyInference("checkpoints/best_model.pt", "data/tokenizer.json")- print("\nGuppy Chat (type 'quit' to exit)")- msgs = []- while True:- inp = input("\nYou> ").strip()- if inp.lower() in ("quit", "exit", "q"):- break- msgs.append({"role": "user", "content": inp})- r = engine.chat_completion(msgs)- msg = r["choices"][0]["message"]- if msg.get("content"):- print(f"Guppy> {msg['content']}")- msgs.append(msg)+ from .inference import main+ main()else:print(f"Unknown command: {cmd}")diff --git a/guppylm/inference.py b/guppylm/inference.pyindex 6bf2976..86a444a 100644--- a/guppylm/inference.py+++ b/guppylm/inference.py@@ -103,9 +103,16 @@ def main():p.add_argument("--checkpoint", default="checkpoints/best_model.pt")p.add_argument("--tokenizer", default="data/tokenizer.json")p.add_argument("--device", default="cpu")+ p.add_argument("--prompt", "-p", help="Single prompt mode: ask one question and exit")args = p.parse_args()engine = GuppyInference(args.checkpoint, args.tokenizer, args.device)++ if args.prompt:+ result = engine.chat_completion([{"role": "user", "content": args.prompt}])+ print(result["choices"][0]["message"]["content"])+ return+print("\nGuppy Chat (type 'quit' to exit)")msgs = []while True: