vLLM: --enable-auto-tool-choice --tool-call-parser lfm2 (OpenCode sends tool_choice=auto)
Browse files- run_job.sh +2 -0
- serve_policy.sh +2 -0
run_job.sh
CHANGED
|
@@ -58,6 +58,8 @@ vllm serve "${MODEL:-LiquidAI/LFM2.5-2.6B}" \
|
|
| 58 |
--max-model-len 32768 \
|
| 59 |
--return-tokens-as-token-ids \
|
| 60 |
--logprobs-mode processed_logprobs \
|
|
|
|
|
|
|
| 61 |
--weight-transfer-config '{"backend":"nccl"}' &
|
| 62 |
VLLM_PID=$!
|
| 63 |
|
|
|
|
| 58 |
--max-model-len 32768 \
|
| 59 |
--return-tokens-as-token-ids \
|
| 60 |
--logprobs-mode processed_logprobs \
|
| 61 |
+
--enable-auto-tool-choice \
|
| 62 |
+
--tool-call-parser lfm2 \
|
| 63 |
--weight-transfer-config '{"backend":"nccl"}' &
|
| 64 |
VLLM_PID=$!
|
| 65 |
|
serve_policy.sh
CHANGED
|
@@ -22,4 +22,6 @@ vllm serve "$MODEL" \
|
|
| 22 |
--max-model-len 32768 \
|
| 23 |
--return-tokens-as-token-ids \
|
| 24 |
--logprobs-mode processed_logprobs \
|
|
|
|
|
|
|
| 25 |
--weight-transfer-config '{"backend":"nccl"}'
|
|
|
|
| 22 |
--max-model-len 32768 \
|
| 23 |
--return-tokens-as-token-ids \
|
| 24 |
--logprobs-mode processed_logprobs \
|
| 25 |
+
--enable-auto-tool-choice \
|
| 26 |
+
--tool-call-parser lfm2 \
|
| 27 |
--weight-transfer-config '{"backend":"nccl"}'
|