Fix(MInference): fix e2e benchmark guideline & fix A-shape multi gpu (#…

…66) Co-authored-by: Yucheng Li <[email protected]> Co-authored-by: Chengruidong Zhang <[email protected]> Co-authored-by: Yuqing Yang <[email protected]>
microsoft · Aug 8, 2024 · f0cae77 · f0cae77
1 parent 7a11a33
commit f0cae77
Show file tree

Hide file tree

Showing 3 changed files with 54 additions and 46 deletions.
diff --git a/experiments/README.md b/experiments/README.md
@@ -51,7 +51,11 @@ wget https://raw.githubusercontent.com/FranxYao/chain-of-thought-hub/main/gsm8k/
 2. Run a single context window size test using one method:
 
 ```bash
-python experiments/benchmarks/benchmark_e2e.py --attn_type hf --context_window 1_000_000
+# If the context window is greater than 700K, you need to enable kv_cache_cpu.
+python experiments/benchmarks/benchmark_e2e.py --attn_type minference --context_window 1_000_000 --kv_cache_cpu
+python experiments/benchmarks/benchmark_e2e.py --attn_type minference_with_dense --context_window 1_000_000 --kv_cache_cpu
+
+python experiments/benchmarks/benchmark_e2e.py --attn_type minference --context_window 500_000
 ```
 
 3. Run all latency experiments using different methods:

diff --git a/experiments/infinite_bench/run_infinitebench.py b/experiments/infinite_bench/run_infinitebench.py
@@ -247,9 +247,11 @@ def load_model(
         if os.path.exists(output_path) and not args.rewrite:
             print(f"Output file {output_path} exists. Loading from file.")
             compute_scores(output_path, data_name, real_model_name, max_seq_length)
+            with open(output_path) as f:
+                preds = [json.loads(ii) for ii in f.readlines()]
 
         for i, eg in tqdm(enumerate(examples)):
-            if i < args.start_example_id:
+            if i < args.start_example_id or i < len(preds):
                 continue
             input_text = create_prompt(eg, data_name, real_model_name, args.data_dir)
             ground_truth = get_answer(eg, data_name)

diff --git a/minference/ops/streaming_kernel.py b/minference/ops/streaming_kernel.py
@@ -429,55 +429,57 @@ def _forward(
     global _BLOCK_M
 
     try:
-        _attn_fwd[grid](
-            q, k, v, sm_scale, m, o, l, #
-            q.stride(0), q.stride(1), q.stride(2), q.stride(3),  #
-            k.stride(0), k.stride(1), k.stride(2), k.stride(3),  #
-            v.stride(0), v.stride(1), v.stride(2), v.stride(3),  #
-            o.stride(0), o.stride(1), o.stride(2), o.stride(3),  #
-            q.shape[0], q.shape[1], k.shape[1], #
-            q.shape[2],  #
-            q_round_len,
-            k.shape[2],
-            sliding_window_offset,
-            sliding_window_size,
-            BLOCK_DMODEL=Lk,  #
-            END=end,
-            INIT=init,
-            BLOCK_M=_BLOCK_M,
-            BLOCK_N=_BLOCK_N,
-            SLIDING_WINDOW=(sliding_window is not None),
-            COMPLEMENT_SLIDING_WINDOW=complement_sliding_window,
-            num_warps=4,
-            num_stages=4
-        )
+        with torch.cuda.device(q.device):
+            _attn_fwd[grid](
+                q, k, v, sm_scale, m, o, l, #
+                q.stride(0), q.stride(1), q.stride(2), q.stride(3),  #
+                k.stride(0), k.stride(1), k.stride(2), k.stride(3),  #
+                v.stride(0), v.stride(1), v.stride(2), v.stride(3),  #
+                o.stride(0), o.stride(1), o.stride(2), o.stride(3),  #
+                q.shape[0], q.shape[1], k.shape[1], #
+                q.shape[2],  #
+                q_round_len,
+                k.shape[2],
+                sliding_window_offset,
+                sliding_window_size,
+                BLOCK_DMODEL=Lk,  #
+                END=end,
+                INIT=init,
+                BLOCK_M=_BLOCK_M,
+                BLOCK_N=_BLOCK_N,
+                SLIDING_WINDOW=(sliding_window is not None),
+                COMPLEMENT_SLIDING_WINDOW=complement_sliding_window,
+                num_warps=4,
+                num_stages=4
+            )
     except triton.OutOfResources as E:
         _BLOCK_N = _BLOCK_N // 2
         _BLOCK_M = _BLOCK_M // 2
         from warnings import warn
         warn(f"Triton Attention Output Resources. {E}\nUse smaller block size {_BLOCK_N}.")
-        _attn_fwd[grid](
-            q, k, v, sm_scale, m, o, l, #
-            q.stride(0), q.stride(1), q.stride(2), q.stride(3),  #
-            k.stride(0), k.stride(1), k.stride(2), k.stride(3),  #
-            v.stride(0), v.stride(1), v.stride(2), v.stride(3),  #
-            o.stride(0), o.stride(1), o.stride(2), o.stride(3),  #
-            q.shape[0], q.shape[1], k.shape[1], #
-            q.shape[2],  #
-            q_round_len,
-            k.shape[2],
-            sliding_window_offset,
-            sliding_window_size,
-            BLOCK_DMODEL=Lk,  #
-            END=end,
-            INIT=init,
-            BLOCK_M=_BLOCK_M,
-            BLOCK_N=_BLOCK_N,
-            SLIDING_WINDOW=(sliding_window is not None),
-            COMPLEMENT_SLIDING_WINDOW=complement_sliding_window,
-            num_warps=4,
-            num_stages=4
-        )
+        with torch.cuda.device(q.device):
+            _attn_fwd[grid](
+                q, k, v, sm_scale, m, o, l, #
+                q.stride(0), q.stride(1), q.stride(2), q.stride(3),  #
+                k.stride(0), k.stride(1), k.stride(2), k.stride(3),  #
+                v.stride(0), v.stride(1), v.stride(2), v.stride(3),  #
+                o.stride(0), o.stride(1), o.stride(2), o.stride(3),  #
+                q.shape[0], q.shape[1], k.shape[1], #
+                q.shape[2],  #
+                q_round_len,
+                k.shape[2],
+                sliding_window_offset,
+                sliding_window_size,
+                BLOCK_DMODEL=Lk,  #
+                END=end,
+                INIT=init,
+                BLOCK_M=_BLOCK_M,
+                BLOCK_N=_BLOCK_N,
+                SLIDING_WINDOW=(sliding_window is not None),
+                COMPLEMENT_SLIDING_WINDOW=complement_sliding_window,
+                num_warps=4,
+                num_stages=4
+            )
 
 
     if end: