@HenryNdubuaku full results below but this is for weights that were loaded from huggingface, just making sure those are weights at the correct precision. Can't easily run reconvert with these models because it causes OOM on raspberry pi, but if hf doesn't have the int4 versions i can work around that.

╔══════════════════════════════════════════╗
║ 1K CONTEXT TEST ║
╚══════════════════════════════════════════╝
Response: The data represents the circumference of circles as they get closer to a specific value (100.530880). The circumference grows by 2 units for each circle added until it reaches 125.66 units, which is the closest value before reaching the next multiple of 100. As each subsequent circle's circumference adds only 1 unit (compared to an average increase of approximately 2 per previous round), they gradually converge on this target without achieving it exactly, demonstrating how small adjustments can be made for [-> stopped]

[Results]
"success": true,
"error": null,
"cloud_handoff": false,
"response": "The data represents the circumference of circles as they get closer to a specific value (100.530880). The circumference grows by 2 units for each circle added until it reaches 125.66 units, which is the closest value before reaching the next multiple of 100. As each subsequent circle's circumference adds only 1 unit (compared to an average increase of approximately 2 per previous round), they gradually converge on this target without achieving it exactly, demonstrating how small adjustments can be made for",
"function_calls": [],
"confidence": 0.8480,
"time_to_first_token_ms": 13129.16,
"total_time_ms": 22427.45,
"prefill_tps": 69.24,
"decode_tps": 10.65,
"ram_usage_mb": 868.52,
"prefill_tokens": 909,
"decode_tokens": 100,
"total_tokens": 1009
└─ Status: PASSED ✓
✓ PASS │ 1k_context

╔══════════════════════════════════════════╗
║ TRANSCRIPTION ║
╚══════════════════════════════════════════╝
Transcript: hello hello hello just quickly testing out creating a wa file through voice memos the goal is to use this wite file to test out whisper hopefully this will transcribe properly that's all i can hope for all right here we go

[Results]
"success": true,
"error": null,
"cloud_handoff": false,
"response": "hello hello hello just quickly testing out creating a wa file through voice memos the goal is to use this wite file to test out whisper hopefully this will transcribe properly that's all i can hope for all right here we go",
"function_calls": [],
"confidence": 1.0000,
"time_to_first_token_ms": 4581.88,
"total_time_ms": 4582.28,
"prefill_tps": 0.00,
"decode_tps": 182044.89,
"ram_usage_mb": 1931.03,
"prefill_tokens": 0,
"decode_tokens": 74,
"total_tokens": 74
✓ PASS │ transcription

╔══════════════════════════════════════════╗
║ VLM MULTI-TURN TEST ║
╚══════════════════════════════════════════╝

[Turn 1]
User: Describe what is happening in this image in two sentences.
Assistant: A close-up image of a brown monkey with dark green eyes, giving an intense gaze into the camera. The animal's face features deep wrinkles around its mouth and nose area. The background consists of blurred leaves, suggesting an outdoor environment in daylight.

[Results - Turn 1]
"success": true,
"error": null,
"cloud_handoff": false,
"response": "A close-up image of a brown monkey with dark green eyes, giving an intense gaze into the camera. The animal's face features deep wrinkles around its mouth and nose area. The background consists of blurred leaves, suggesting an outdoor environment in daylight.",
"function_calls": [],
"confidence": 0.8523,
"time_to_first_token_ms": 8910.40,
"total_time_ms": 13364.53,
"prefill_tps": 9.76,
"decode_tps": 11.23,
"ram_usage_mb": 2107.34,
"prefill_tokens": 87,
"decode_tokens": 51,
"total_tokens": 138

[Turn 2]
User: Describe the image once again.
Assistant: Here we have an image of a brown monkey with dark green eyes, capturing its intense gaze into the camera with wrinkled face features and soft fur adorning its body. The blurred green background indicates a natural environment where this animal is situated indoors or in captivity, creating the illusion of it being outside but looking very close up inside.

[Results - Turn 2]
"success": true,
"error": null,
"cloud_handoff": false,
"response": "Here we have an image of a brown monkey with dark green eyes, capturing its intense gaze into the camera with wrinkled face features and soft fur adorning its body. The blurred green background indicates a natural environment where this animal is situated indoors or in captivity, creating the illusion of it being outside but looking very close up inside.",
"function_calls": [],
"confidence": 0.8612,
"time_to_first_token_ms": 9347.62,
"total_time_ms": 15411.22,
"prefill_tps": 16.58,
"decode_tps": 10.88,
"ram_usage_mb": 2320.09,
"prefill_tokens": 155,
"decode_tokens": 67,
"total_tokens": 222
✓ PASS │ vlm_multiturn
────────────────────────────────────────────────────────────────────────────────────────
✓ All 1 tests passed!