update qwen conf

conf/model/llm/qwen/qwen2-VL-2B-instruct.ini

@@ -0,0 +1,46 @@
+[LLAMA3]
+llm_model_path="../weights/llm/qwen/Qwen2-VL-2B-Instruct/qwen2-vl-2B-instruct-Q4_K_M.gguf"
+mmproj_model_path="../weights/llm/qwen/Qwen2-VL-2B-Instruct/qwen2-vl-2B-instruct-vision-fp16.gguf"
+n_gpu_layers=80
+main_gpu_device=0
+vocab_only=false
+
+[SAMPLER]
+# 0 = disabled, otherwise samplers should return at least min_keep tokens
+min_keep=1
+# <= 0 to use vocab size
+top_k=100
+# 1.0 = disabled
+top_p=0.95
+# 0.0 = disabled
+min_p=0.05
+# 0.0 = disabled
+xtc_probability=0.0
+# > 0.5 disables XTC
+xtc_threshold=0.10
+# typical_p, 1.0 = disabled
+typ_p=1.00
+# <= 0.0 to sample greedily, 0.0 to not output probabilities
+temp=0.8
+# 1.0 = disabled
+penalty_repeat= 1.00
+# 0.0 = disabled
+penalty_freq= 0.00
+# 0.0 = disabled
+penalty_present= 0.00
+# 0.0 = disabled;
+dry_multiplier= 0.0
+# 0.0 = disabled
+dry_base=1.75
+# tokens extending repetitions beyond this receive penalty
+dry_allowed_length=2
+# how many tokens to scan for repetitions (0 = disable penalty, -1 = context size)
+dry_penalty_last_n=-1
+# consider newlines as a repeatable token
+penalize_nl=false
+ignore_eos=false
+# disable performance metrics
+no_perf=false
+
+[CONTEXT]
+context_size=8192