Spaces:
Running
Running
Update .env.example to include TOKEN for gated Hugging Face repos and modify app.py to support TOKEN or HF_TOKEN for loading tokenizers.
Browse files- .env.example +3 -2
- app.py +11 -2
.env.example
CHANGED
|
@@ -1,2 +1,3 @@
|
|
| 1 |
-
|
| 2 |
-
|
|
|
|
|
|
| 1 |
+
# For gated Hugging Face repos (e.g. some LLaMA, Bielik). Use one of:
|
| 2 |
+
TOKEN=your_huggingface_token_here
|
| 3 |
+
# HF_TOKEN=your_huggingface_token_here
|
app.py
CHANGED
|
@@ -3,10 +3,14 @@ Tokenizer Playground — inspect how text is tokenized, view config and chat tem
|
|
| 3 |
Runs locally and on Hugging Face Spaces.
|
| 4 |
"""
|
| 5 |
import html
|
|
|
|
| 6 |
import random
|
| 7 |
import gradio as gr
|
| 8 |
from transformers import AutoTokenizer
|
| 9 |
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
def _token_color(index: int, total: int) -> str:
|
| 12 |
"""Distinct hue per token, pastel background (HSL)."""
|
|
@@ -99,11 +103,16 @@ def get_tokenizer(model_id: str, use_fast: bool = True):
|
|
| 99 |
key = (model_id or DEFAULT_MODEL, use_fast)
|
| 100 |
if key not in _tokenizer_cache:
|
| 101 |
try:
|
| 102 |
-
|
| 103 |
-
model_id or DEFAULT_MODEL,
|
| 104 |
trust_remote_code=True,
|
| 105 |
use_fast=use_fast,
|
| 106 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
except Exception as e:
|
| 108 |
raise RuntimeError(f"Failed to load tokenizer for '{model_id}': {e}") from e
|
| 109 |
return _tokenizer_cache[key]
|
|
|
|
| 3 |
Runs locally and on Hugging Face Spaces.
|
| 4 |
"""
|
| 5 |
import html
|
| 6 |
+
import os
|
| 7 |
import random
|
| 8 |
import gradio as gr
|
| 9 |
from transformers import AutoTokenizer
|
| 10 |
|
| 11 |
+
# For gated repos: set TOKEN or HF_TOKEN (e.g. in .env or Hugging Face Space secrets)
|
| 12 |
+
HF_TOKEN = os.getenv("TOKEN") or os.getenv("HF_TOKEN") or os.getenv("HUGGING_FACE_HUB_TOKEN") or None
|
| 13 |
+
|
| 14 |
|
| 15 |
def _token_color(index: int, total: int) -> str:
|
| 16 |
"""Distinct hue per token, pastel background (HSL)."""
|
|
|
|
| 103 |
key = (model_id or DEFAULT_MODEL, use_fast)
|
| 104 |
if key not in _tokenizer_cache:
|
| 105 |
try:
|
| 106 |
+
kwargs = dict(
|
|
|
|
| 107 |
trust_remote_code=True,
|
| 108 |
use_fast=use_fast,
|
| 109 |
)
|
| 110 |
+
if HF_TOKEN:
|
| 111 |
+
kwargs["token"] = HF_TOKEN
|
| 112 |
+
_tokenizer_cache[key] = AutoTokenizer.from_pretrained(
|
| 113 |
+
model_id or DEFAULT_MODEL,
|
| 114 |
+
**kwargs,
|
| 115 |
+
)
|
| 116 |
except Exception as e:
|
| 117 |
raise RuntimeError(f"Failed to load tokenizer for '{model_id}': {e}") from e
|
| 118 |
return _tokenizer_cache[key]
|