tuieval 0.2.0.dev42__py3-none-any.whl → 0.2.0.dev43__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- tuieval/_version.py +2 -2
- tuieval/client.py +11 -1
- tuieval/engine.py +3 -0
- {tuieval-0.2.0.dev42.dist-info → tuieval-0.2.0.dev43.dist-info}/METADATA +1 -1
- {tuieval-0.2.0.dev42.dist-info → tuieval-0.2.0.dev43.dist-info}/RECORD +8 -8
- {tuieval-0.2.0.dev42.dist-info → tuieval-0.2.0.dev43.dist-info}/WHEEL +0 -0
- {tuieval-0.2.0.dev42.dist-info → tuieval-0.2.0.dev43.dist-info}/entry_points.txt +0 -0
- {tuieval-0.2.0.dev42.dist-info → tuieval-0.2.0.dev43.dist-info}/licenses/LICENSE +0 -0
tuieval/_version.py
CHANGED
|
@@ -18,7 +18,7 @@ version_tuple: tuple[int | str, ...]
|
|
|
18
18
|
commit_id: str | None
|
|
19
19
|
__commit_id__: str | None
|
|
20
20
|
|
|
21
|
-
__version__ = version = '0.2.0.
|
|
22
|
-
__version_tuple__ = version_tuple = (0, 2, 0, '
|
|
21
|
+
__version__ = version = '0.2.0.dev43'
|
|
22
|
+
__version_tuple__ = version_tuple = (0, 2, 0, 'dev43')
|
|
23
23
|
|
|
24
24
|
__commit_id__ = commit_id = None
|
tuieval/client.py
CHANGED
|
@@ -17,6 +17,10 @@ import time
|
|
|
17
17
|
import urllib.parse
|
|
18
18
|
|
|
19
19
|
SERVER_ERROR_PREFIXES = ("server returned HTTP 5", "server returned HTTP 429", "server error", "connection error")
|
|
20
|
+
# the model can't be reached there at all (a bad key, no credits, or a hosted API with no provider serving it, e.g.
|
|
21
|
+
# OpenRouter's 404 "No endpoints found"): not the model's answer either, and waiting won't fix it
|
|
22
|
+
UNAVAILABLE_PREFIXES = ("server returned HTTP 401", "server returned HTTP 402", "server returned HTTP 403",
|
|
23
|
+
"server returned HTTP 404")
|
|
20
24
|
|
|
21
25
|
|
|
22
26
|
def is_server_error(error):
|
|
@@ -27,9 +31,15 @@ def is_server_error(error):
|
|
|
27
31
|
and "timeout" not in error.lower()
|
|
28
32
|
|
|
29
33
|
|
|
34
|
+
def is_unavailable(error):
|
|
35
|
+
"""True for an error saying the model isn't available on that server at all: HTTP 401, 402, 403 or
|
|
36
|
+
404. Never the model's answer, and not worth retrying."""
|
|
37
|
+
return bool(error) and error.startswith(UNAVAILABLE_PREFIXES)
|
|
38
|
+
|
|
39
|
+
|
|
30
40
|
def server_error_row(rec):
|
|
31
41
|
"""A stored answer that was really a server failure (never a judgement of the model)."""
|
|
32
|
-
return not rec.get("pass") and is_server_error(rec.get("reason"))
|
|
42
|
+
return not rec.get("pass") and (is_server_error(rec.get("reason")) or is_unavailable(rec.get("reason")))
|
|
33
43
|
|
|
34
44
|
|
|
35
45
|
class Cancelled(Exception):
|
tuieval/engine.py
CHANGED
|
@@ -2105,6 +2105,9 @@ class Engine:
|
|
|
2105
2105
|
self._check()
|
|
2106
2106
|
if res["error"] and not alive():
|
|
2107
2107
|
raise ModelFailed(f"the server stopped: {res['error']}")
|
|
2108
|
+
if client.is_unavailable(res["error"]): # never an answer; retrying won't help
|
|
2109
|
+
raise ModelFailed(f"the model isn't available there (not counted against it): "
|
|
2110
|
+
f"{res['error'][:300]}")
|
|
2108
2111
|
if not client.is_server_error(res["error"]):
|
|
2109
2112
|
break
|
|
2110
2113
|
# the server failed, not the model: never recorded as an answer
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.5
|
|
2
2
|
Name: tuieval
|
|
3
|
-
Version: 0.2.0.
|
|
3
|
+
Version: 0.2.0.dev43
|
|
4
4
|
Summary: Evaluate local and frontier LLMs on your own questions: accuracy, speed, tokens and PASS/FAIL verdicts, in the terminal.
|
|
5
5
|
Project-URL: Homepage, https://github.com/ashe-wb/tuieval
|
|
6
6
|
Project-URL: Issues, https://github.com/ashe-wb/tuieval/issues
|
|
@@ -1,12 +1,12 @@
|
|
|
1
1
|
tuieval/__init__.py,sha256=myVXVX7htbwezW6k3x_FuKgn1RSZ2WcTaDEMByGn6Wg,312
|
|
2
2
|
tuieval/__main__.py,sha256=E6Gls0DNz8GQK2K-kOUIx8cYhgANW_CH54VKrfCfs14,52
|
|
3
|
-
tuieval/_version.py,sha256=
|
|
3
|
+
tuieval/_version.py,sha256=5PoXfeeUc_RJHZEcbrdVeWQpyK7qC0e8oom2QPT307Y,535
|
|
4
4
|
tuieval/builtin.toml,sha256=Kxw22RZCJhURUyu5Wz8tj5Y1nM85A8hFVRbA4ocoPUg,4293
|
|
5
5
|
tuieval/cli.py,sha256=kI1P0a7TMS-daHjZjLDBdBDPOwBCmPw1qjw6VDlY258,4505
|
|
6
|
-
tuieval/client.py,sha256=
|
|
6
|
+
tuieval/client.py,sha256=kpsyLEZKfxCsvIGBHBbBy83TpI97CvnGCDq-4AbJS4Y,9986
|
|
7
7
|
tuieval/compare.py,sha256=SHN_o_nlppFcHMBqBZRITIkeQDPaR_JxhUZItQLMBwo,26270
|
|
8
8
|
tuieval/doctor.py,sha256=y1C_ggjd8oikXTrovgTBf26wjThP8EoVkfTVeO_n5dQ,6070
|
|
9
|
-
tuieval/engine.py,sha256=
|
|
9
|
+
tuieval/engine.py,sha256=w3KsgxzFAMcRyVzFGQXhTgN3d9qYRtAZZxNLXZOgnrQ,114222
|
|
10
10
|
tuieval/export.py,sha256=dbHQgL_ubWtIPM1sLzAew_gETnrDJyd0zgqVA5vkpXM,9657
|
|
11
11
|
tuieval/machines.py,sha256=ImEihNj8QlBRdCHLED4QIdjJUvjVtZgo6DiVU5h6Zzs,13415
|
|
12
12
|
tuieval/onboard.py,sha256=IYETVDWCCo9vF0Dtj4_rr9VanYeCky8PwP3-2g5A_l0,6215
|
|
@@ -46,8 +46,8 @@ tuieval/templates/packs/tool_call/pack.toml,sha256=n4bx0lGJHSmW41uVqhfIXJ2BWZpmT
|
|
|
46
46
|
tuieval/templates/packs/tool_call/system.txt,sha256=mKxu-lB-0YjLHXdgeFJa8x2C180OvcNz5cHv2N5Bg3U,254
|
|
47
47
|
tuieval/templates/packs/tool_call/tests.yaml,sha256=Nv7c4syrzJ9cU8EBKSK_1ZSFx9V02aCjW4-2TBCPT7s,1859
|
|
48
48
|
tuieval/templates/packs/tool_call/tools.yaml,sha256=HgxCkh0On2VwbHSsjAGM8tNCK7ZtKLGrs7lMCn3Fprs,1048
|
|
49
|
-
tuieval-0.2.0.
|
|
50
|
-
tuieval-0.2.0.
|
|
51
|
-
tuieval-0.2.0.
|
|
52
|
-
tuieval-0.2.0.
|
|
53
|
-
tuieval-0.2.0.
|
|
49
|
+
tuieval-0.2.0.dev43.dist-info/METADATA,sha256=P6oHoB_GI8wBLV-_DUqU4TggqZdTVGtz5glAfpIn2jU,13881
|
|
50
|
+
tuieval-0.2.0.dev43.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
|
|
51
|
+
tuieval-0.2.0.dev43.dist-info/entry_points.txt,sha256=7eIhi1GpFxBP2OLJV02Nn8LwsNuO512eKOGt1IBbj04,45
|
|
52
|
+
tuieval-0.2.0.dev43.dist-info/licenses/LICENSE,sha256=7wfRgGtBGH1aXhZkIn47oUznqKC18kEPMvUVgol7Ffc,1077
|
|
53
|
+
tuieval-0.2.0.dev43.dist-info/RECORD,,
|
|
File without changes
|
|
File without changes
|
|
File without changes
|