Merge pull request #56 from VE-FORBRYDERNE/concurrency

Compile TPU backend in background

aiserver.py

@@ -7,10 +7,10 @@
 
 # External packages
 import eventlet
-eventlet.monkey_patch()
+eventlet.monkey_patch(all=True, thread=False)
 import os
 os.system("")
-os.environ['EVENTLET_THREADPOOL_SIZE'] = '1'
+os.environ['EVENTLET_THREADPOOL_SIZE'] = '50'
 from eventlet import tpool
 
 from os import path, getcwd
@@ -21,6 +21,7 @@ import zipfile
 import packaging
 import contextlib
 import traceback
+import threading
 from typing import Any, Callable, TypeVar, Union, Dict, Set, List
 
 import requests
@@ -976,6 +977,29 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly", "TPUMeshTransforme
         from transformers import GPT2TokenizerFast
         tokenizer = GPT2TokenizerFast.from_pretrained("gpt2", cache_dir="cache/")
 else:
+    def tpumtjgetsofttokens():
+        soft_tokens = None
+        if(vars.sp is None):
+            global np
+            if 'np' not in globals():
+                import numpy as np
+            tensor = np.zeros((1, tpu_mtj_backend.params["d_model"]), dtype=np.float32)
+            rows = tensor.shape[0]
+            padding_amount = tpu_mtj_backend.params["seq"] - (tpu_mtj_backend.params["seq"] % -tpu_mtj_backend.params["cores_per_replica"]) - rows
+            tensor = np.pad(tensor, ((0, padding_amount), (0, 0)))
+            tensor = tensor.reshape(
+                tpu_mtj_backend.params["cores_per_replica"],
+                -1,
+                tpu_mtj_backend.params["d_model"],
+            )
+            vars.sp = tensor
+        soft_tokens = np.arange(
+            tpu_mtj_backend.params["n_vocab"] + tpu_mtj_backend.params["n_vocab_padding"],
+            tpu_mtj_backend.params["n_vocab"] + tpu_mtj_backend.params["n_vocab_padding"] + vars.sp_length,
+            dtype=np.uint32
+        )
+        return soft_tokens
+
     # If we're running Colab or OAI, we still need a tokenizer.
     if(vars.model == "Colab"):
         from transformers import GPT2TokenizerFast
@@ -992,6 +1016,17 @@ else:
         vars.allowsp = True
         vars.modeldim = int(tpu_mtj_backend.params["d_model"])
         tokenizer = tpu_mtj_backend.tokenizer
+        soft_tokens = tpumtjgetsofttokens()
+        threading.Thread(  # Compile backend code in background
+            target=tpu_mtj_backend.infer,
+            args=(np.uint32((23403, 727, 20185)),),
+            kwargs={
+                "soft_embeddings": vars.sp,
+                "soft_tokens": soft_tokens,
+                "gen_len": 1,
+                "numseqs": vars.numseqs,
+            },
+        ).start()
 
 # Set up Flask routes
 @app.route('/')
@@ -1583,7 +1618,8 @@ def execute_outmod():
 #  Lua runtime startup
 #==================================================================#
 
-print(colors.PURPLE + "Initializing Lua Bridge... " + colors.END, end="")
+print("", end="", flush=True)
+print(colors.PURPLE + "Initializing Lua Bridge... " + colors.END, end="", flush=True)
 
 # Set up Lua state
 vars.lua_state = lupa.LuaRuntime(unpack_returned_tuples=True)
@@ -2863,27 +2899,8 @@ def tpumtjgenerate(txt, minimum, maximum, found_entries=None):
     try:
         if(vars.dynamicscan):
             raise ValueError("Dynamic world info scanning is not supported by the TPU backend yet")
-        
+
-        soft_tokens = None
+        soft_tokens = tpumtjgetsofttokens()
-        if(vars.sp is None):
-            global np
-            if 'np' not in globals():
-                import numpy as np
-            tensor = np.zeros((1, tpu_mtj_backend.params["d_model"]), dtype=np.float32)
-            rows = tensor.shape[0]
-            padding_amount = tpu_mtj_backend.params["seq"] - (tpu_mtj_backend.params["seq"] % -tpu_mtj_backend.params["cores_per_replica"]) - rows
-            tensor = np.pad(tensor, ((0, padding_amount), (0, 0)))
-            tensor = tensor.reshape(
-                tpu_mtj_backend.params["cores_per_replica"],
-                -1,
-                tpu_mtj_backend.params["d_model"],
-            )
-            vars.sp = tensor
-        soft_tokens = np.arange(
-            tpu_mtj_backend.params["n_vocab"] + tpu_mtj_backend.params["n_vocab_padding"],
-            tpu_mtj_backend.params["n_vocab"] + tpu_mtj_backend.params["n_vocab_padding"] + vars.sp_length,
-            dtype=np.uint32
-        )
 
         genout = tpool.execute(
             tpu_mtj_backend.infer,
@@ -4335,8 +4352,9 @@ loadsettings()
 #==================================================================#
 #  Final startup commands to launch Flask app
 #==================================================================#
+print("", end="", flush=True)
 if __name__ == "__main__":
-    print("{0}\nStarting webserver...{1}".format(colors.GREEN, colors.END))
+    print("{0}\nStarting webserver...{1}".format(colors.GREEN, colors.END), flush=True)
 
     # Start Flask/SocketIO (Blocking, so this must be last method!)
     
@@ -4361,4 +4379,4 @@ if __name__ == "__main__":
         socketio.run(app, port=5000)
 
 else:
-    print("{0}\nServer started in WSGI mode!{1}".format(colors.GREEN, colors.END))
+    print("{0}\nServer started in WSGI mode!{1}".format(colors.GREEN, colors.END), flush=True)