From a6c4298d27f740117d92476dc7fa10df8cd70014 Mon Sep 17 00:00:00 2001
From: Nikita Shulga <2453524+malfet@users.noreply.github.com>
Date: Tue, 14 May 2024 09:24:29 -0700
Subject: [PATCH] Fix browser init (#797)

Update prompt chat is waiting for, which was modified by https://github.com/pytorch/torchchat/pull/476

Modify logging defaults to not create a file in a temp folder without prompting user, but rather just print an info messages Replace few `prints` with `logging.info` This way, information about bandwith achieved will be printed to the console, but not to the web-browser chat window

Test plan:
```
% python3 torchchat.py browser stories110M &
% curl -L http://127.0.0.1:5000
% curl -d "prompt=Once upon a time" -X POST http://127.0.0.1:5000/chat
```

TODOs:
- Add CI that repeats above steps
-Figure out if spawning generator from the browser can be avoided

Fixes https://github.com/pytorch/torchchat/issues/785
---
 chat_in_browser.py | 4 +++-
 cli.py             | 5 +----
 generate.py        | 6 +++---
 3 files changed, 7 insertions(+), 8 deletions(-)

diff --git a/chat_in_browser.py b/chat_in_browser.py
index f012ab105..e835fa009 100644
--- a/chat_in_browser.py
+++ b/chat_in_browser.py
@@ -35,7 +35,7 @@ def main():
             except:
                 continue
 
-            if decoded.startswith("System Prompt") and decoded.endswith(": "):
+            if decoded.endswith("Do you want to enter a system prompt? Enter y for yes and anything else for no. \n"):
                 print(f"| {decoded}")
                 proc.stdin.write("\n".encode("utf-8"))
                 proc.stdin.flush()
@@ -93,6 +93,8 @@ def chat():
         model_prefix = "Model: "
         if output.startswith(model_prefix):
             output = output[len(model_prefix) :]
+        else:
+            print("But output is", output)
 
         global convo
 
diff --git a/cli.py b/cli.py
index 5223cf3ce..a28a98863 100644
--- a/cli.py
+++ b/cli.py
@@ -15,10 +15,7 @@
 from build.utils import allowable_dtype_names, allowable_params_table, get_device_str
 from download import download_and_convert, is_model_downloaded
 
-FORMAT = (
-    "%(levelname)s: %(asctime)-15s: %(filename)s: %(funcName)s: %(module)s: %(message)s"
-)
-logging.basicConfig(filename="/tmp/torchchat.log", level=logging.INFO, format=FORMAT)
+logging.basicConfig(level=logging.INFO,format="%(message)s")
 logger = logging.getLogger(__name__)
 
 default_device = os.getenv("TORCHCHAT_DEVICE", "fast")
diff --git a/generate.py b/generate.py
index b7e7c105f..81bcffaee 100644
--- a/generate.py
+++ b/generate.py
@@ -752,12 +752,12 @@ def callback(x):
             # Don't continue here.... because we need to report and reset
             # continue
 
-        print(
+        logging.info(
             f"Time for inference {i + 1}: {t:.02f} sec total, {tokens_sec:.02f} tokens/sec"
         )
-        print(f"Bandwidth achieved: {model_size * tokens_sec / 1e9:.02f} GB/s")
+        logging.info(f"Bandwidth achieved: {model_size * tokens_sec / 1e9:.02f} GB/s")
         if i == 0:
-            print(
+            logging.info(
                 f"*** This first iteration will include cold start effects for dynamic import, hardware caches{', JIT compilation' if jit_compile else ''}. ***"
             )
         if start_pos >= max_seq_length: