biondizzle/deepseek-v4-quant
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add commit hashes to run history table

Browse Source
This commit is contained in:
biondizzle
2026-05-09 06:47:26 +00:00
parent d7593fc1dd
commit 9438af5a8c
1 changed files with 7 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
README.md
View File

@@ -59,11 +59,14 @@ python3 /root/nvidia-meeting/deepseek-v4-quant/scripts/quantize_nvfp4.py --valid
## Run History (forward progression)
| Run | Date | Script | Calib | Result | Root Cause | Fix |
| Run | Date | Commit | Calib | Result | Root Cause | Fix |
|-----|------|--------|-------|--------|------------|-----|
| 1 | May 7 | shell wrapper, FP8 source | 256 | ❌ Crashed at batch probing | `o_b_proj` shape mismatch — finegrained_fp8 wraps MLA projections incorrectly with FP8 source | Use BF16 source (dequantized) |
| 2 | May 8-9 | shell wrapper, BF16 source | 128 | ❌ Crashed at export (128/128 calib ✅) | `get_activation_scaling_factor` reads stale GPU amax → CUDA illegal memory access | Snapshot amax to CPU immediately after calibration |
| 3 | May 9 | `quantize_nvfp4.py` v1 | 128 | 🔄 Running | — | — |
| 1 | May 7 | shell wrapper (pre-repo) | 256 | ❌ Crashed at batch probing | `o_b_proj` shape mismatch — finegrained_fp8 wraps MLA projections incorrectly with FP8 source | Use BF16 source (dequantized) |
| 2 | May 8-9 | shell wrapper (pre-repo) | 128 | ❌ Crashed at export (128/128 calib ✅) | `get_activation_scaling_factor` reads stale GPU amax → CUDA illegal memory access | Snapshot amax to CPU immediately after calibration |
| 3 | May 9 | `3907838` | 128 | 🔄 Running | — | — |
**If Run 3 succeeds**, the v2 script (`6eeba26`) is reference only — no need to adopt it.
**If Run 3 fails at export**, use commit `6eaba26` (adds amax snapshot + force CPU + validate mode).
**Key lesson from Run 2:** The `use_seq_device_map` mode shuffles layers through GPU for calibration. Quantizer amax tensors sit in VRAM for 5+ hours while CUDA's allocator churns memory. By export time, the GPU tensor metadata is valid but the underlying memory has been recycled — reading it triggers `cudaErrorIllegalAddress`. The fix is to copy amax to CPU *immediately* after calibration, before any further GPU operations.