mudassiruddin21 commited on
Commit
f0048ae
·
verified ·
1 Parent(s): 5fd0620

Results of the first run

Browse files
Files changed (1) hide show
  1. README.md +37 -9
README.md CHANGED
@@ -71,9 +71,9 @@ before choosing it:
71
  * **Objective — the engine's drafting loop.** `emcdraft.head.unroll` runs all three draft steps
72
  for every position of a sequence in parallel, with exactly the key/value visibility the
73
  engine's draft cache has (verified prefix, then the round's own entries). Each step is
74
- trained on the token the target really emitted; step 1 also distils the target's full
75
- next-token distribution; later steps are weighted by whether the earlier ones were right,
76
- because the engine only reaches them in that case.
77
  * **FP8 with QAT.** An exact, function-preserving re-scaling (`equalize_`) first moves the big
78
  matrices into the range where E4M3 has relative precision; they are then trained through
79
  E4M3 rounding with a straight-through estimator, so the compact file is the model that was
@@ -87,8 +87,29 @@ before choosing it:
87
 
88
  ## Results
89
 
90
- Produced by `scripts/pipeline.sh`; raw kit outputs and logs are in `results/` and in the
91
- dataset repo. *(Filled in by the run — see `results/summary.md`.)*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92
 
93
  ## Layout
94
 
@@ -113,7 +134,7 @@ tests/test_head.py CPU self-test (FP8 codec, unroll vs. step-by-step emulatio
113
 
114
  ## Reproduce
115
 
116
- Linux, one GPU with at least 80 GB, about 110 GB of disk. In a Python shell or notebook on
117
  that machine:
118
 
119
  ```python
@@ -139,9 +160,16 @@ This starts `scripts/supervise.py`, which runs `scripts/pipeline.sh` stage by st
139
  | eval_main, eval_baseline, eval_alt | full official `eval` (160 dev prompts) of the draft, the baseline and the second variant |
140
 
141
  Every stage resumes after an interruption, and logs, kit results, generated data and the
142
- draft folders are pushed to the dataset repo as they are produced. The engine is the
143
- pinned version installed from PyPI rather than the container image; the results JSON records
144
- this. `python tests/test_head.py` runs the CPU self-test (no GPU, no downloads).
 
 
 
 
 
 
 
145
 
146
  ## Submission format and rules
147
 
 
71
  * **Objective — the engine's drafting loop.** `emcdraft.head.unroll` runs all three draft steps
72
  for every position of a sequence in parallel, with exactly the key/value visibility the
73
  engine's draft cache has (verified prefix, then the round's own entries). Each step is
74
+ trained on the token the target really emitted and also distils the target's full
75
+ distribution for that token; later steps are weighted by whether the earlier ones were
76
+ right, because the engine only reaches them in that case.
77
  * **FP8 with QAT.** An exact, function-preserving re-scaling (`equalize_`) first moves the big
78
  matrices into the range where E4M3 has relative precision; they are then trained through
79
  E4M3 rounding with a straight-through estimator, so the compact file is the model that was
 
87
 
88
  ## Results
89
 
90
+ Official kit `eval` (vLLM 0.31.0, all 160 dev prompts, K = 3), every row measured in the same
91
+ run on the same **NVIDIA RTX PRO 6000 Blackwell**:
92
+
93
+ | draft | τ | chat / code / math / long-form | TPOT | memory | score, same GPU | score, published baseline |
94
+ |---|---|---|---|---|---|---|
95
+ | baseline (the target's built-in MTP head) | 2.806 | 3.06 / 3.74 / 3.64 / 2.37 | 15.39 ms | 0.917 GB | 100.0 | 96.2 |
96
+ | this draft, bf16 | 2.885 | 3.07 / 3.75 / 3.70 / 2.48 | 15.31 ms | 0.917 GB | 102.3 | 98.5 |
97
+ | **this draft, FP8 (submitted variant)** | **2.868** | 3.08 / 3.78 / 3.68 / 2.45 | 15.27 ms | **0.607 GB** | **114.4** | **110.7** |
98
+
99
+ * *Score, same GPU* uses the baseline as measured above; *score, published baseline* is the
100
+ kit's own number, whose speed term compares against the organisers' TPOT from different
101
+ hardware (that is why the baseline itself gets 96.2 there).
102
+ * The FP8 variant's same-GPU score splits into acceptance 1.034 (×0.5), memory 1.5 (×0.25,
103
+ the cap) and speed 1.008 (×0.25). Most of the gain is the compact storage; fine-tuning added
104
+ +0.06 to τ, mostly in long-form writing (+0.08) and math (+0.04).
105
+ * The offline simulator predicted the dev acceptance within 0.015 of the engine (2.858 vs
106
+ 2.868 for the FP8 draft, 2.792 vs 2.806 for the baseline).
107
+ * Run r1 trained on 2.02 M generated tokens (4,820 answers) for 584 steps (13 min); its
108
+ validation τ was still rising when the schedule ended. Total GPU time for the whole pipeline,
109
+ including the three full evaluations: about 2.6 hours.
110
+
111
+ Raw kit outputs, per-stage logs, machine statistics and the training summary are in the
112
+ dataset repo under `runs/r1/`; the exported draft folders under `submissions/r1/`.
113
 
114
  ## Layout
115
 
 
134
 
135
  ## Reproduce
136
 
137
+ Linux, one GPU with at least 80 GB, about 120 GB of disk. In a Python shell or notebook on
138
  that machine:
139
 
140
  ```python
 
160
  | eval_main, eval_baseline, eval_alt | full official `eval` (160 dev prompts) of the draft, the baseline and the second variant |
161
 
162
  Every stage resumes after an interruption, and logs, kit results, generated data and the
163
+ draft folders are pushed to the dataset repo as they are produced. Settings in
164
+ `control/run.env` extend a finished run instead of repeating it: `GEN_APPEND=1` and
165
+ `EXTRACT_APPEND=1` keep earlier answers and features, `TRAIN_INIT` starts from an earlier
166
+ head, `RUN_NAME` names the new training run.
167
+
168
+ The engine is the pinned version installed from PyPI rather than the container image; the
169
+ results JSON records this. The pipeline sets `VLLM_USE_FLASHINFER_SAMPLER=0`, so the engine
170
+ uses its built-in sampler instead of compiling FlashInfer's at start-up; decoding is greedy,
171
+ so outputs and draft behaviour are unaffected. `python tests/test_head.py` runs the CPU
172
+ self-test (no GPU, no downloads).
173
 
174
  ## Submission format and rules
175