feat(intern-decision): stack, DNS and GPU 3 acceptance for the SemIf replacement

stacks/intern-decision: compose (GPU 1, :8033, hard VRAM cap as the single .env knob, healthcheck, Homepage group 'AI - Eval & Retrieval'), .env.example and README. dns: intern-decision.fv.internal -> fv-ml1 (synced to ana/esh/nh3). acceptance on fv-ml1 GPU 3, 3 fresh processes: bit-identical to the Jev bench's native rows (pooled 240/259, Wyrd 79/84, 0/560 flips, Δp 0), negative control 10/122/14, 0 flips across restarts; largest accepted request 200 at a 10,134 MiB card peak under a 9.25 GiB cap; 503 and recovery proven at a tight cap. GPU 1 deploy held: nvidia-smi Free on GPU 1 is 15,442 MiB.
2026-09-30 09:38:00 -07:00
parent 21d16d7ad8
commit a262477a61
59 changed files with 11969 additions and 1 deletions
@@ -0,0 +1 @@
+1d98f12c909d19f7e157823e29412e4fb44aa22db94c6ee4a96d8453aed29c70
@@ -0,0 +1,3 @@
+1790785010.413756011 sets:start
+1790785010.470560236 sets:done
+gpu3 after: 2 MiB
@@ -0,0 +1 @@
+{"status":"ok","model":{"name":"Intern-Decision-4B","source":"internlm/Intern-Decision-4B","revision":"0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd","checkpoint":"/hf/hub/models--internlm--Intern-Decision-4B/snapshots/0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd","inference_py_sha256":"c904e2c67ca0775621a22375ee373d2ba30b52117cda870c6c9ef74143b29863","temperature":1.99241824,"dtype":"bfloat16","attn_implementation":"sdpa","device":"cuda","max_length":8192,"torch_version":"2.10.0+cu128","transformers_version":"5.17.0","vision_tower":"removed","device_name":"NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition","allocated_gib":7.937,"reserved_gib":7.969,"max_reserved_gib":8.861},"vram_cap_gib":9.25,"max_tokens":8192,"max_decisions":64,"max_questions_per_call":16,"chunking":"/decide/shared questions are packed greedily, in request order, into calls of at most 16 (1-16, 17-32, ...); each call is one prompt, so the questions in a call are asked together. With orderings, ordering k of every decision forms wave k, packed the same way.","workloads":[]}
@@ -0,0 +1 @@
+{"status":"ok","model":{"name":"Intern-Decision-4B","source":"internlm/Intern-Decision-4B","revision":"0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd","checkpoint":"/hf/hub/models--internlm--Intern-Decision-4B/snapshots/0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd","inference_py_sha256":"c904e2c67ca0775621a22375ee373d2ba30b52117cda870c6c9ef74143b29863","temperature":1.99241824,"dtype":"bfloat16","attn_implementation":"sdpa","device":"cuda","max_length":8192,"torch_version":"2.10.0+cu128","transformers_version":"5.17.0","vision_tower":"removed","device_name":"NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition","allocated_gib":7.937,"reserved_gib":7.969,"max_reserved_gib":8.861},"vram_cap_gib":9.25,"max_tokens":8192,"max_decisions":64,"max_questions_per_call":16,"chunking":"/decide/shared questions are packed greedily, in request order, into calls of at most 16 (1-16, 17-32, ...); each call is one prompt, so the questions in a call are asked together. With orderings, ordering k of every decision forms wave k, packed the same way.","workloads":[]}
@@ -0,0 +1,28 @@
+# pid=1840754 gpu=GPU-186dacf4-f447-ef03-ffc3-12ab28e1b8ff
+1790785007.592863455,8812,8822
+1790785007.701879593,8812,8822
+1790785007.808902409,8812,8822
+1790785007.915876585,8812,8822
+1790785008.021279853,8812,8822
+1790785008.129614494,8812,8822
+1790785008.237274588,8812,8822
+1790785008.347694802,8812,8822
+1790785008.453754747,8812,8822
+1790785008.563765228,8812,8822
+1790785008.672580344,8812,8822
+1790785008.781543962,8812,8822
+1790785008.889253036,8812,8822
+1790785008.995532764,8812,8822
+1790785009.100480397,8812,8822
+1790785009.209746018,8812,8822
+1790785009.316264048,8812,8822
+1790785009.423762660,8812,8822
+1790785009.532005020,8812,8822
+1790785009.643907242,8812,8822
+1790785009.752750529,8812,8822
+1790785009.861883069,8812,8822
+1790785009.969250889,8812,8822
+1790785010.076693700,8812,8822
+1790785010.185418326,8812,8822
+1790785010.293297732,8812,8822
+1790785010.399472568,8812,8822
				`@@ -0,0 +1 @@`
				`1d98f12c909d19f7e157823e29412e4fb44aa22db94c6ee4a96d8453aed29c70`
				`@@ -0,0 +1 @@`
				{"status":"ok","model":{"name":"Intern-Decision-4B","source":"internlm/Intern-Decision-4B","revision":"0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd","checkpoint":"/hf/hub/models--internlm--Intern-Decision-4B/snapshots/0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd","inference_py_sha256":"c904e2c67ca0775621a22375ee373d2ba30b52117cda870c6c9ef74143b29863","temperature":1.99241824,"dtype":"bfloat16","attn_implementation":"sdpa","device":"cuda","max_length":8192,"torch_version":"2.10.0+cu128","transformers_version":"5.17.0","vision_tower":"removed","device_name":"NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition","allocated_gib":7.937,"reserved_gib":7.969,"max_reserved_gib":8.861},"vram_cap_gib":9.25,"max_tokens":8192,"max_decisions":64,"max_questions_per_call":16,"chunking":"/decide/shared questions are packed greedily, in request order, into calls of at most 16 (1-16, 17-32, ...); each call is one prompt, so the questions in a call are asked together. With orderings, ordering k of every decision forms wave k, packed the same way.","workloads":[]}