[Mini-VLLM] Cut non-OpenAI surfaces, dead kernels, Mistral stack
- Remove Anthropic, Cohere, SageMaker, decisions, structured-decisions, rl, render-launcher and batch entrypoints; keep scale-out token routes - Drop the Swagger static bundle and the /docs, /redoc routes - Delete device-keyed tuning configs for fused_moe, quant utils, mamba select-state and helion; no sm86 entry ever existed for any of them - Remove ten unused rope variants plus the fbgemm_fp8 and fp_quant quantizers and their config, metrics and loader plumbing - Cut the CPU device build: csrc/cpu, cpu_extension.cmake, cpu requirements and the setup.py cpu target path - Delete 29 environment variables nothing reads and 14 chat-template fallbacks for models this fork no longer registers - Remove nine dead CUDA ops (deepseek_v4, kimi_k3, minimax, muse_spark, dsv3 gemm) with their kernels, bindings, wrappers and CMake entries - Delete the connector test suites that targeted removed connectors - Cut the Mistral tokenizer and parser stack, its tests, and the mistral_common and ijson dependency pins - Restore the Qwen4Exp registry rows and nvidia/mtp.py that an earlier cut deleted while keeping the model tree Co-authored-by: Hinata <hinata@idihore.id> Signed-off-by: MOVZX <movzx@yahoo.com>
This commit is contained in:
@@ -122,20 +122,20 @@ def _track(model: nn.Module, loaded: set[str], quantized: bool) -> str:
|
||||
|
||||
def test_weights_track_reports_missing_quantized_weights():
|
||||
# Quantized checkpoints may only omit KV-cache quantization params, empty
|
||||
# placeholders, fbgemm_fp8's input_scale_ub and online-quant params.
|
||||
# placeholders, config-built input_scale_ub and online-quant params.
|
||||
online = _QuantMethod()
|
||||
online.uses_meta_device = True
|
||||
model = nn.Module()
|
||||
model.proj = _layer(_QuantMethod(), weight=(4, 4), weight_scale_inv=(1, 1))
|
||||
model.unquantized = _layer(UnquantizedLinearMethod(), weight=(4, 4), bias=(4,))
|
||||
model.gptq = _layer(_QuantMethod(), qweight=(4, 4), qzeros=(0,))
|
||||
model.fbgemm = _layer(_QuantMethod(), weight=(4, 4), input_scale_ub=())
|
||||
model.cfg_built = _layer(_QuantMethod(), weight=(4, 4), input_scale_ub=())
|
||||
model.attn = _layer(BaseKVCacheMethod(None), k_scale=(), k_zero_point=())
|
||||
model.online = _layer(online, weight=(4, 4))
|
||||
loaded = {"proj.weight", "unquantized.weight", "gptq.qweight", "fbgemm.weight"}
|
||||
loaded = {"proj.weight", "unquantized.weight", "gptq.qweight", "cfg_built.weight"}
|
||||
error = _track(model, loaded, quantized=True)
|
||||
assert "proj.weight_scale_inv" in error and "unquantized.bias" in error
|
||||
assert not any(n in error for n in ("gptq", "fbgemm", "attn", "online"))
|
||||
assert not any(n in error for n in ("gptq", "cfg_built", "attn", "online"))
|
||||
|
||||
|
||||
def test_weights_track_keeps_unquantized_models_unchanged():
|
||||
|
||||
Reference in New Issue
Block a user