MrA7A3 commited on
Commit
ec0b9da
·
verified ·
1 Parent(s): cf81ae1

KAPO transformers runtime rollout

Browse files
.env CHANGED
@@ -14,14 +14,15 @@ HF_SPACE_DOCKER=1
14
  KAGGLE_AUTO_BOOTSTRAP=0
15
  KAPO_BOOTSTRAP_URL=https://drive.google.com/uc?export=download&id=19jyBWsQ9ciJVPi2PUigu5ti3gJ24A6TG
16
  KAPO_COMPUTE_PROFILE=cpu
17
- KAPO_HF_INFERENCE_API=1
18
- KAPO_HF_TRANSFORMERS_RUNTIME=0
 
19
  KAPO_LAZY_EMBED_STARTUP=1
20
  KAPO_LAZY_MODEL_STARTUP=1
21
  KAPO_PATCH_BUNDLE_URL=https://drive.google.com/uc?export=download&id=16rIe05GZihhAz7ba8E-WibKaJKbh9eu1
22
  KAPO_PATCH_MANIFEST_URL=https://drive.google.com/uc?export=download&id=1jLuPMCA3hp9qstZZtpBNzTK0XOmLrV8b
23
- KAPO_REMOTE_ENV_PASSWORD_B64=Wml4VURXdlRyU2FXQ011czNxODF5dEdVNmZ1Ujh3VmY
24
- KAPO_REMOTE_ENV_URL_B64=aHR0cHM6Ly9kcml2ZS5nb29nbGUuY29tL3VjP2V4cG9ydD1kb3dubG9hZCZpZD0xYlhYeWJaQUpqb3lQeGIxNUJDUXhRYmpJb3VQRHh4TU4
25
  KAPO_SHARED_STATE_BACKEND=google_drive
26
  MODEL_PROFILE_ID=hf-debugger-qwen25-7b-instruct
27
  MODEL_REPO=Qwen/Qwen2.5-1.5B-Instruct
 
14
  KAGGLE_AUTO_BOOTSTRAP=0
15
  KAPO_BOOTSTRAP_URL=https://drive.google.com/uc?export=download&id=19jyBWsQ9ciJVPi2PUigu5ti3gJ24A6TG
16
  KAPO_COMPUTE_PROFILE=cpu
17
+ KAPO_DISABLE_EMBEDDINGS=1
18
+ KAPO_HF_INFERENCE_API=0
19
+ KAPO_HF_TRANSFORMERS_RUNTIME=1
20
  KAPO_LAZY_EMBED_STARTUP=1
21
  KAPO_LAZY_MODEL_STARTUP=1
22
  KAPO_PATCH_BUNDLE_URL=https://drive.google.com/uc?export=download&id=16rIe05GZihhAz7ba8E-WibKaJKbh9eu1
23
  KAPO_PATCH_MANIFEST_URL=https://drive.google.com/uc?export=download&id=1jLuPMCA3hp9qstZZtpBNzTK0XOmLrV8b
24
+ KAPO_REMOTE_ENV_PASSWORD_B64=RDlxTjFLNU42ZDVpZURDYjBrRUVOV1hONFN5OVVnZUM
25
+ KAPO_REMOTE_ENV_URL_B64=aHR0cHM6Ly9kcml2ZS5nb29nbGUuY29tL3VjP2V4cG9ydD1kb3dubG9hZCZpZD0xTTcySjRWeVFqVlNESUFTblZMLXF2dEZCa3VRM1R1WkQ
26
  KAPO_SHARED_STATE_BACKEND=google_drive
27
  MODEL_PROFILE_ID=hf-debugger-qwen25-7b-instruct
28
  MODEL_REPO=Qwen/Qwen2.5-1.5B-Instruct
bootstrap_space_runtime.py CHANGED
@@ -22,14 +22,15 @@ DEFAULT_ENV = {
22
  "KAGGLE_AUTO_BOOTSTRAP": "0",
23
  "KAPO_BOOTSTRAP_URL": "https://drive.google.com/uc?export=download&id=19jyBWsQ9ciJVPi2PUigu5ti3gJ24A6TG",
24
  "KAPO_COMPUTE_PROFILE": "cpu",
25
- "KAPO_HF_INFERENCE_API": "1",
26
- "KAPO_HF_TRANSFORMERS_RUNTIME": "0",
 
27
  "KAPO_LAZY_EMBED_STARTUP": "1",
28
  "KAPO_LAZY_MODEL_STARTUP": "1",
29
  "KAPO_PATCH_BUNDLE_URL": "https://drive.google.com/uc?export=download&id=16rIe05GZihhAz7ba8E-WibKaJKbh9eu1",
30
  "KAPO_PATCH_MANIFEST_URL": "https://drive.google.com/uc?export=download&id=1jLuPMCA3hp9qstZZtpBNzTK0XOmLrV8b",
31
- "KAPO_REMOTE_ENV_PASSWORD_B64": "Wml4VURXdlRyU2FXQ011czNxODF5dEdVNmZ1Ujh3VmY",
32
- "KAPO_REMOTE_ENV_URL_B64": "aHR0cHM6Ly9kcml2ZS5nb29nbGUuY29tL3VjP2V4cG9ydD1kb3dubG9hZCZpZD0xYlhYeWJaQUpqb3lQeGIxNUJDUXhRYmpJb3VQRHh4TU4",
33
  "KAPO_SHARED_STATE_BACKEND": "google_drive",
34
  "MODEL_PROFILE_ID": "hf-debugger-qwen25-7b-instruct",
35
  "MODEL_REPO": "Qwen/Qwen2.5-1.5B-Instruct",
 
22
  "KAGGLE_AUTO_BOOTSTRAP": "0",
23
  "KAPO_BOOTSTRAP_URL": "https://drive.google.com/uc?export=download&id=19jyBWsQ9ciJVPi2PUigu5ti3gJ24A6TG",
24
  "KAPO_COMPUTE_PROFILE": "cpu",
25
+ "KAPO_DISABLE_EMBEDDINGS": "1",
26
+ "KAPO_HF_INFERENCE_API": "0",
27
+ "KAPO_HF_TRANSFORMERS_RUNTIME": "1",
28
  "KAPO_LAZY_EMBED_STARTUP": "1",
29
  "KAPO_LAZY_MODEL_STARTUP": "1",
30
  "KAPO_PATCH_BUNDLE_URL": "https://drive.google.com/uc?export=download&id=16rIe05GZihhAz7ba8E-WibKaJKbh9eu1",
31
  "KAPO_PATCH_MANIFEST_URL": "https://drive.google.com/uc?export=download&id=1jLuPMCA3hp9qstZZtpBNzTK0XOmLrV8b",
32
+ "KAPO_REMOTE_ENV_PASSWORD_B64": "RDlxTjFLNU42ZDVpZURDYjBrRUVOV1hONFN5OVVnZUM",
33
+ "KAPO_REMOTE_ENV_URL_B64": "aHR0cHM6Ly9kcml2ZS5nb29nbGUuY29tL3VjP2V4cG9ydD1kb3dubG9hZCZpZD0xTTcySjRWeVFqVlNESUFTblZMLXF2dEZCa3VRM1R1WkQ",
34
  "KAPO_SHARED_STATE_BACKEND": "google_drive",
35
  "MODEL_PROFILE_ID": "hf-debugger-qwen25-7b-instruct",
36
  "MODEL_REPO": "Qwen/Qwen2.5-1.5B-Instruct",
brain_server/api/main.py CHANGED
@@ -118,6 +118,14 @@ def _feature_enabled(name: str, default: bool = False) -> bool:
118
  return str(value).strip().lower() in {"1", "true", "yes", "on"}
119
 
120
 
 
 
 
 
 
 
 
 
121
  def _remote_brain_only() -> bool:
122
  return _feature_enabled("REMOTE_BRAIN_ONLY", default=False)
123
 
@@ -258,6 +266,9 @@ def ensure_model_loaded(repo_id: str, filename: str, hf_token: str | None = None
258
 
259
  def _load_embed_model() -> None:
260
  global EMBED_MODEL
 
 
 
261
  if EMBED_MODEL is not None:
262
  return
263
 
@@ -273,6 +284,9 @@ def _load_default_model() -> None:
273
  repo_id = str(os.getenv("MODEL_REPO", DEFAULT_MODEL_REPO) or DEFAULT_MODEL_REPO).strip()
274
  filename = str(os.getenv("MODEL_FILE", "") or "").strip()
275
  provider = str(os.getenv("BRAIN_PROVIDER", "") or os.getenv("BRAIN_TEMPLATE", "") or "").strip().lower()
 
 
 
276
  if not filename and (_feature_enabled("KAPO_HF_INFERENCE_API", default=False) or "huggingface" in provider or "hf-space" in provider):
277
  MODEL = None
278
  MODEL_ERROR = None
 
118
  return str(value).strip().lower() in {"1", "true", "yes", "on"}
119
 
120
 
121
+ def _hf_transformers_runtime_enabled() -> bool:
122
+ return _feature_enabled("KAPO_HF_TRANSFORMERS_RUNTIME", default=False)
123
+
124
+
125
+ def _embeddings_enabled() -> bool:
126
+ return not _feature_enabled("KAPO_DISABLE_EMBEDDINGS", default=False)
127
+
128
+
129
  def _remote_brain_only() -> bool:
130
  return _feature_enabled("REMOTE_BRAIN_ONLY", default=False)
131
 
 
266
 
267
  def _load_embed_model() -> None:
268
  global EMBED_MODEL
269
+ if not _embeddings_enabled():
270
+ logger.info("Embedding model disabled by configuration")
271
+ return
272
  if EMBED_MODEL is not None:
273
  return
274
 
 
284
  repo_id = str(os.getenv("MODEL_REPO", DEFAULT_MODEL_REPO) or DEFAULT_MODEL_REPO).strip()
285
  filename = str(os.getenv("MODEL_FILE", "") or "").strip()
286
  provider = str(os.getenv("BRAIN_PROVIDER", "") or os.getenv("BRAIN_TEMPLATE", "") or "").strip().lower()
287
+ if _hf_transformers_runtime_enabled():
288
+ ensure_model_loaded(repo_id, filename, hf_token=os.getenv("HF_TOKEN"))
289
+ return
290
  if not filename and (_feature_enabled("KAPO_HF_INFERENCE_API", default=False) or "huggingface" in provider or "hf-space" in provider):
291
  MODEL = None
292
  MODEL_ERROR = None
kapo.env CHANGED
@@ -14,14 +14,15 @@ HF_SPACE_DOCKER=1
14
  KAGGLE_AUTO_BOOTSTRAP=0
15
  KAPO_BOOTSTRAP_URL=https://drive.google.com/uc?export=download&id=19jyBWsQ9ciJVPi2PUigu5ti3gJ24A6TG
16
  KAPO_COMPUTE_PROFILE=cpu
17
- KAPO_HF_INFERENCE_API=1
18
- KAPO_HF_TRANSFORMERS_RUNTIME=0
 
19
  KAPO_LAZY_EMBED_STARTUP=1
20
  KAPO_LAZY_MODEL_STARTUP=1
21
  KAPO_PATCH_BUNDLE_URL=https://drive.google.com/uc?export=download&id=16rIe05GZihhAz7ba8E-WibKaJKbh9eu1
22
  KAPO_PATCH_MANIFEST_URL=https://drive.google.com/uc?export=download&id=1jLuPMCA3hp9qstZZtpBNzTK0XOmLrV8b
23
- KAPO_REMOTE_ENV_PASSWORD_B64=Wml4VURXdlRyU2FXQ011czNxODF5dEdVNmZ1Ujh3VmY
24
- KAPO_REMOTE_ENV_URL_B64=aHR0cHM6Ly9kcml2ZS5nb29nbGUuY29tL3VjP2V4cG9ydD1kb3dubG9hZCZpZD0xYlhYeWJaQUpqb3lQeGIxNUJDUXhRYmpJb3VQRHh4TU4
25
  KAPO_SHARED_STATE_BACKEND=google_drive
26
  MODEL_PROFILE_ID=hf-debugger-qwen25-7b-instruct
27
  MODEL_REPO=Qwen/Qwen2.5-1.5B-Instruct
 
14
  KAGGLE_AUTO_BOOTSTRAP=0
15
  KAPO_BOOTSTRAP_URL=https://drive.google.com/uc?export=download&id=19jyBWsQ9ciJVPi2PUigu5ti3gJ24A6TG
16
  KAPO_COMPUTE_PROFILE=cpu
17
+ KAPO_DISABLE_EMBEDDINGS=1
18
+ KAPO_HF_INFERENCE_API=0
19
+ KAPO_HF_TRANSFORMERS_RUNTIME=1
20
  KAPO_LAZY_EMBED_STARTUP=1
21
  KAPO_LAZY_MODEL_STARTUP=1
22
  KAPO_PATCH_BUNDLE_URL=https://drive.google.com/uc?export=download&id=16rIe05GZihhAz7ba8E-WibKaJKbh9eu1
23
  KAPO_PATCH_MANIFEST_URL=https://drive.google.com/uc?export=download&id=1jLuPMCA3hp9qstZZtpBNzTK0XOmLrV8b
24
+ KAPO_REMOTE_ENV_PASSWORD_B64=RDlxTjFLNU42ZDVpZURDYjBrRUVOV1hONFN5OVVnZUM
25
+ KAPO_REMOTE_ENV_URL_B64=aHR0cHM6Ly9kcml2ZS5nb29nbGUuY29tL3VjP2V4cG9ydD1kb3dubG9hZCZpZD0xTTcySjRWeVFqVlNESUFTblZMLXF2dEZCa3VRM1R1WkQ
26
  KAPO_SHARED_STATE_BACKEND=google_drive
27
  MODEL_PROFILE_ID=hf-debugger-qwen25-7b-instruct
28
  MODEL_REPO=Qwen/Qwen2.5-1.5B-Instruct
requirements.txt CHANGED
@@ -7,4 +7,7 @@ requests==2.32.3
7
  python-json-logger==2.0.7
8
  cryptography>=43.0.0,<46.0.0
9
  huggingface_hub>=0.33.5,<2.0
 
 
 
10
  starlette>=0.40.0,<1.0
 
7
  python-json-logger==2.0.7
8
  cryptography>=43.0.0,<46.0.0
9
  huggingface_hub>=0.33.5,<2.0
10
+ torch>=2.2,<3.0
11
+ transformers==4.57.6
12
+ accelerate>=1.0,<2.0
13
  starlette>=0.40.0,<1.0