compile cache model card update
Browse files- compile_cache_model_card.md +10 -10
compile_cache_model_card.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
-
# Compile Cache
|
| 2 |
|
| 3 |
-
Generated: 2026-05-
|
| 4 |
|
| 5 |
## Purpose
|
| 6 |
Records the exact configuration under which the compile cache (inductor + Triton) was built.
|
|
@@ -17,7 +17,7 @@ against these values to pinpoint the cache-key mismatch.
|
|
| 17 |
|---------|---------|
|
| 18 |
| Torch | 2.6.0+rocm6.2.4 |
|
| 19 |
| HIP / ROCm | 6.2.41134-65d174c3e |
|
| 20 |
-
| Triton | 3.
|
| 21 |
| ck4inductor | 0.0.11 |
|
| 22 |
| Python | 3.10.12 |
|
| 23 |
|
|
@@ -35,15 +35,15 @@ against these values to pinpoint the cache-key mismatch.
|
|
| 35 |
|
| 36 |
## Model
|
| 37 |
- Architecture: EnhancedDSRModel (~1.743B total params)
|
| 38 |
-
- Trainable params: fp32 (~337M
|
| 39 |
-
- Frozen backbone: bf16 (~1.406B
|
| 40 |
- Sequence length: 1024 tokens
|
| 41 |
|
| 42 |
## Cache Statistics
|
| 43 |
| Cache | Files |
|
| 44 |
|-------|-------|
|
| 45 |
-
| inductor\_cache |
|
| 46 |
-
| triton\_cache |
|
| 47 |
| tar.gz size | 110.3 MB |
|
| 48 |
|
| 49 |
## Autotune Cache Key
|
|
@@ -58,6 +58,6 @@ If AUTOTUNE blocks appear on a pod that should have cache hits, verify:
|
|
| 58 |
2. `TORCHINDUCTOR_CK_DIR` points to valid CK root: `/home/hotaisle/.local/lib/python3.10/site-packages/ck4inductor`
|
| 59 |
3. `library/src` symlink inside ck4inductor package exists (needed for gemm_universal kernel enumeration)
|
| 60 |
4. `TORCHINDUCTOR_CACHE_DIR` == `/tmp/ml_cache/inductor_cache` (with extracted cache files present)
|
| 61 |
-
5. `TRITON_CACHE_DIR` == `/tmp/ml_cache/triton_cache` (must NOT be `~/.triton`
|
| 62 |
-
6. Torch version matches: `2.6.0+rocm6.2.4`
|
| 63 |
-
7. Triton version matches: `3.
|
|
|
|
| 1 |
+
# Compile Cache — MI300X
|
| 2 |
|
| 3 |
+
Generated: 2026-05-07T17:26:34Z
|
| 4 |
|
| 5 |
## Purpose
|
| 6 |
Records the exact configuration under which the compile cache (inductor + Triton) was built.
|
|
|
|
| 17 |
|---------|---------|
|
| 18 |
| Torch | 2.6.0+rocm6.2.4 |
|
| 19 |
| HIP / ROCm | 6.2.41134-65d174c3e |
|
| 20 |
+
| Triton | 3.2.0 |
|
| 21 |
| ck4inductor | 0.0.11 |
|
| 22 |
| Python | 3.10.12 |
|
| 23 |
|
|
|
|
| 35 |
|
| 36 |
## Model
|
| 37 |
- Architecture: EnhancedDSRModel (~1.743B total params)
|
| 38 |
+
- Trainable params: fp32 (~337M — new alignment modules)
|
| 39 |
+
- Frozen backbone: bf16 (~1.406B — RWKV-7 1.5B World)
|
| 40 |
- Sequence length: 1024 tokens
|
| 41 |
|
| 42 |
## Cache Statistics
|
| 43 |
| Cache | Files |
|
| 44 |
|-------|-------|
|
| 45 |
+
| inductor\_cache | 1019 |
|
| 46 |
+
| triton\_cache | 8202 |
|
| 47 |
| tar.gz size | 110.3 MB |
|
| 48 |
|
| 49 |
## Autotune Cache Key
|
|
|
|
| 58 |
2. `TORCHINDUCTOR_CK_DIR` points to valid CK root: `/home/hotaisle/.local/lib/python3.10/site-packages/ck4inductor`
|
| 59 |
3. `library/src` symlink inside ck4inductor package exists (needed for gemm_universal kernel enumeration)
|
| 60 |
4. `TORCHINDUCTOR_CACHE_DIR` == `/tmp/ml_cache/inductor_cache` (with extracted cache files present)
|
| 61 |
+
5. `TRITON_CACHE_DIR` == `/tmp/ml_cache/triton_cache` (must NOT be `~/.triton` — that is wiped on restart)
|
| 62 |
+
6. Torch version matches: `2.6.0+rocm6.2.4` — different minor/patch versions can invalidate kernel hashes
|
| 63 |
+
7. Triton version matches: `3.2.0`
|