File size: 3,344 Bytes
8efb28e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 | import pytest
from utils import *
server = ServerPreset.tinyllama2()
@pytest.fixture(autouse=True)
def create_server():
global server
server = ServerPreset.tinyllama2()
server.slot_save_path = "./tmp"
server.temperature = 0.0
def test_slot_save_restore():
global server
server.start()
# First prompt in slot 1 should be fully processed
res = server.make_request("POST", "/completion", data={
"prompt": "What is the capital of France?",
"id_slot": 1,
"cache_prompt": True,
})
assert res.status_code == 200
assert match_regex("(Whiskers|Flana)+", res.body["content"])
assert res.body["timings"]["prompt_n"] == 21 # all tokens are processed
# Save state of slot 1
res = server.make_request("POST", "/slots/1?action=save", data={
"filename": "slot1.bin",
})
assert res.status_code == 200
assert res.body["n_saved"] == 84
# Since we have cache, this should only process the last tokens
res = server.make_request("POST", "/completion", data={
"prompt": "What is the capital of Germany?",
"id_slot": 1,
"cache_prompt": True,
})
assert res.status_code == 200
assert match_regex("(Jack|said)+", res.body["content"])
assert res.body["timings"]["prompt_n"] == 6 # only different part is processed
# Loading the saved cache into slot 0
res = server.make_request("POST", "/slots/0?action=restore", data={
"filename": "slot1.bin",
})
assert res.status_code == 200
assert res.body["n_restored"] == 84
# Since we have cache, slot 0 should only process the last tokens
res = server.make_request("POST", "/completion", data={
"prompt": "What is the capital of Germany?",
"id_slot": 0,
"cache_prompt": True,
})
assert res.status_code == 200
assert match_regex("(Jack|said)+", res.body["content"])
assert res.body["timings"]["prompt_n"] == 6 # only different part is processed
# For verification that slot 1 was not corrupted during slot 0 load, same thing should work
res = server.make_request("POST", "/completion", data={
"prompt": "What is the capital of Germany?",
"id_slot": 1,
"cache_prompt": True,
})
assert res.status_code == 200
assert match_regex("(Jack|said)+", res.body["content"])
assert res.body["timings"]["prompt_n"] == 1
def test_slot_erase():
global server
server.start()
res = server.make_request("POST", "/completion", data={
"prompt": "What is the capital of France?",
"id_slot": 1,
"cache_prompt": True,
})
assert res.status_code == 200
assert match_regex("(Whiskers|Flana)+", res.body["content"])
assert res.body["timings"]["prompt_n"] == 21 # all tokens are processed
# erase slot 1
res = server.make_request("POST", "/slots/1?action=erase")
assert res.status_code == 200
# re-run the same prompt, it should process all tokens again
res = server.make_request("POST", "/completion", data={
"prompt": "What is the capital of France?",
"id_slot": 1,
"cache_prompt": True,
})
assert res.status_code == 200
assert match_regex("(Whiskers|Flana)+", res.body["content"])
assert res.body["timings"]["prompt_n"] == 21 # all tokens are processed
|