ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge moe-lora-target-fix CI fixes

Browse source 
Merged latest main, resolved _utils.py and KTO test conflicts
This commit is contained in:
Datta Nimmaturi 2026-06-08 20:20:00 +05:30 • committed by GitHub
commit 6f27ecc66e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 156 additions and 28 deletions
Download patch file Download diff file Expand all files Collapse all files

1
.github/workflows/consolidated-tests-ci.yml vendored
View file

@ -990,6 +990,7 @@ jobs:
# First seen on transformers >=5,<6; each represents a slow
# or recursive source-rewriter path the zoo can address.
"beit": "TimeoutError: compile exceeds per-model budget",
"deepseek_ocr2": "TimeoutError: compile exceeds per-model budget",
"sam": "TimeoutError: compile exceeds per-model budget",
"sam_hq": "TimeoutError: compile exceeds per-model budget",
"deepseek_ocr2": "TimeoutError: compile exceeds per-model budget",

32
.github/workflows/studio-inference-smoke.yml vendored
View file

@ -20,7 +20,7 @@
# enable_tools / enabled_tools, and enable_thinking on/off.
#
# 3. JSON, images
# gemma-4-E2B-it UD-IQ3_XXS (~2.4 GiB) + mmproj-F16 (~986 MiB).
# Qwen3-VL-2B-Instruct UD-IQ2_XXS (~570 MiB) + mmproj-F16 (~780 MiB).
# response_format JSON-schema decoding and OpenAI image_url
# (data URI) plus Anthropic source/base64 image inputs.
#
@ -791,9 +791,9 @@ jobs:
runs-on: ubuntu-latest
timeout-minutes: 30
env:
GGUF_REPO: unsloth/gemma-4-E2B-it-GGUF
GGUF_VARIANT: UD-IQ3_XXS
GGUF_FILE: gemma-4-E2B-it-UD-IQ3_XXS.gguf
GGUF_REPO: unsloth/Qwen3-VL-2B-Instruct-GGUF
GGUF_VARIANT: UD-IQ2_XXS
GGUF_FILE: Qwen3-VL-2B-Instruct-UD-IQ2_XXS.gguf
MMPROJ_FILE: mmproj-F16.gguf
STUDIO_PORT: '18890'
HF_HOME: ${{ github.workspace }}/hf-cache
@ -888,13 +888,23 @@ jobs:
-H 'content-type: application/json' \
-d "{\"username\":\"unsloth\",\"password\":\"$NEW\"}" | jq -r .access_token)
echo "API_KEY=$TOKEN" >> "$GITHUB_ENV"
# Load the GGUF (mmproj is auto-detected via the HF repo
# lookup, the cached file is pulled out of HF_HOME).
curl -fs -X POST "http://127.0.0.1:${STUDIO_PORT}/api/inference/load" \
-H "Authorization: Bearer $TOKEN" -H 'content-type: application/json' \
--max-time 900 \
-d "{\"model_path\":\"$GGUF_REPO\",\"gguf_variant\":\"$GGUF_VARIANT\",\"is_lora\":false,\"max_seq_length\":2048}" \
| jq '{status, display_name, is_vision}'
# Retry: llama-server startup can race process teardown after a
# failed attempt. Keep curl out of a pipe so HTTP failures are not
# masked by jq.
LOAD_OK=0
for attempt in 1 2 3; do
HTTP=$(curl -s -o /tmp/load.json -w '%{http_code}' \
-X POST "http://127.0.0.1:${STUDIO_PORT}/api/inference/load" \
-H "Authorization: Bearer $TOKEN" -H 'content-type: application/json' \
--max-time 900 \
-d "{\"model_path\":\"$GGUF_REPO\",\"gguf_variant\":\"$GGUF_VARIANT\",\"is_lora\":false,\"max_seq_length\":2048}")
if [ "$HTTP" = "200" ]; then LOAD_OK=1; break; fi
echo "::warning::/api/inference/load attempt $attempt returned $HTTP; response:"
cat /tmp/load.json || true
sleep 10
done
[ "$LOAD_OK" = "1" ] || { echo "::error::/api/inference/load failed 3 attempts"; exit 22; }
jq '{status, display_name, is_vision}' /tmp/load.json
- name: JSON schema decoding + image input
env:

10
.github/workflows/studio-windows-inference-smoke.yml vendored
View file

@ -13,7 +13,7 @@
# 2. Tool calling Tests
# Qwen3.5-2B UD-Q4_K_XL (~890 MiB).
# 3. JSON, images
# gemma-4-E2B-it UD-Q4_K_XL + mmproj-F16 (~3.4 GiB total).
# Qwen3-VL-2B-Instruct UD-IQ2_XXS + mmproj-F16 (~1.4 GiB total).
# Within the 14 GB windows-latest SSD budget.
name: Windows Studio GGUF CI
@ -843,9 +843,9 @@ jobs:
run:
shell: bash
env:
GGUF_REPO: unsloth/gemma-4-E2B-it-GGUF
GGUF_VARIANT: UD-Q4_K_XL
GGUF_FILE: gemma-4-E2B-it-UD-Q4_K_XL.gguf
GGUF_REPO: unsloth/Qwen3-VL-2B-Instruct-GGUF
GGUF_VARIANT: UD-IQ2_XXS
GGUF_FILE: Qwen3-VL-2B-Instruct-UD-IQ2_XXS.gguf
MMPROJ_FILE: mmproj-F16.gguf
STUDIO_PORT: '18899'
HF_HOME: ${{ github.workspace }}/hf-cache
@ -1123,7 +1123,7 @@ jobs:
)
data_uri = f"data:image/png;base64,{PNG_64X64_RED_B64}"
# On Windows + the gemma-4-E2B mmproj, llama.cpp's vision
# On Windows + the Qwen3-VL mmproj, llama.cpp's vision
# path runs on CPU (no Metal involvement). The wrapper is
# kept for resilience but the vision path is expected to
# work on Windows; an exception here is a real regression.