flint-8bit / validation.json
prince-canuma's picture
Add verified 8bit MLX conversion of RecursiveSI/flint
e065193 verified
Raw History Blame Contribute Delete
3.22 kB
{
"format": "8bit",
"strict_reload": true,
"finite_prefill_decode_logits": true,
"cache_offsets_correct": true,
"source_sidecars_identical": true,
"native_thinking_template_preserved": true,
"temperature": 0.0,
"max_tokens": 512,
"smoke_tests": [
{
"task": "arithmetic",
"prompt": "What is 17 multiplied by 23? Give the answer.",
"text": "<think>We need answer simple multiplication. Need maybe just answer. 17*23=391. Ensure no extra? User asks \"What is 17 multiplied by 23? Give the answer.\" So final maybe 391.</think>391",
"answer": "391",
"checks": {
"expected_391": true
},
"generated_tokens": 48,
"seconds": 0.5520099161658436
},
{
"task": "retrieval",
"prompt": "The passcode is 941731. Repeat the passcode exactly.",
"text": "<think>The user is asking me to repeat a passcode. This is a simple request. I should just repeat it exactly as given.\n\nThe passcode is 941731.\n\nI should repeat it exactly.</think>941731",
"answer": "941731",
"checks": {
"expected_passcode": true
},
"generated_tokens": 43,
"seconds": 0.49429816706106067
},
{
"task": "code",
"prompt": "Write a Python function add(a, b) that returns a + b.",
"text": "<think>We need to write a Python function add(a, b) that returns a + b. This is straightforward. The user likely expects a simple implementation.\n\nWe should provide the code, maybe with a brief explanation.\n\nThe instruction: \"Write a Python function add(a, b) that returns a + b.\"\n\nSo we can just output:\n\ndef add(a, b):\n return a + b\n\nThat's it.\n\nBut maybe they want a full response with code block. I'll provide the function definition and maybe a quick test.\n\nSince it's a simple task, I'll just give the code.</think>Here's a simple Python function that adds two numbers:\n\n```python\ndef add(a, b):\n return a + b\n```\n\nYou can use it like this:\n\n```python\nresult = add(3, 5) # result will be 8\n```",
"answer": "Here's a simple Python function that adds two numbers:\n\n```python\ndef add(a, b):\n return a + b\n```\n\nYou can use it like this:\n\n```python\nresult = add(3, 5) # result will be 8\n```",
"checks": {
"add_function": true
},
"generated_tokens": 173,
"seconds": 1.8621667921543121
}
],
"all_smoke_checks_pass": true,
"device": {
"device_name": "Apple M5 Max",
"max_recommended_working_set_size": 115448725504,
"memory_size": 137438953472,
"architecture": "applegpu_g17s",
"max_buffer_length": 86586540032,
"resource_limit": 499000
},
"mlx_version": "0.32.4",
"mlx_vlm_version": "0.7.10",
"converter_revision": "6d9ba328abbd6129c7461eda86280aa9fbfd9986",
"python": "3.12.15",
"peak_active_gb": 42.345716936,
"seconds": 10.483687666011974,
"swap_before": "vm.swapusage: total = 10240.00M used = 8736.50M free = 1503.50M (encrypted)",
"swap_after": "vm.swapusage: total = 10240.00M used = 8736.50M free = 1503.50M (encrypted)",
"limits": "Three synthetic smoke prompts and finite/cache checks; not a quality benchmark or proof of parity to BF16/Transformers."
}