charieswei
/
TradingAgents
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat: llama.cpp working with TradingAgents

This commit is contained in:
Ivan lee 2025-08-20 12:12:51 +08:00
parent d40c0dd000
commit 1eda6ce34a
4 changed files with 14 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
main.py
View File

@ -10,10 +10,10 @@ config = DEFAULT_CONFIG.copy()
config["llm_provider"] = "llamacpp" # Use a different model config["llm_provider"] = "llamacpp" # Use a different model
config["backend_url"] = "http://localhost:8080/v1" # Use a different backend config["backend_url"] = "http://localhost:8080/v1" # Use a different backend
config["deep_think_llm"] = ( config["deep_think_llm"] = (
"models/Qwen3-4B-Thinking-2507-UD-Q8_K_XL.gguf" # Use a different model "models/Llama-3.3-70B-Instruct.Q5_K_M.gguf" # Use a different model
) )
config["quick_think_llm"] = ( config["quick_think_llm"] = (
"models/Qwen3-4B-Thinking-2507-UD-Q8_K_XL.gguf" # Use a different model "models/Llama-3.3-70B-Instruct.Q5_K_M.gguf" # Use a different model
) )
config["max_debate_rounds"] = 1 # Increase debate rounds config["max_debate_rounds"] = 1 # Increase debate rounds
config["online_tools"] = True # Increase debate rounds config["online_tools"] = True # Increase debate rounds

10
main.slurm
View File

@ -1,13 +1,13 @@
#!/bin/bash #!/bin/bash
#SBATCH --job-name=tradingAgents #SBATCH --job-name=tradingAgents
#SBATCH --output=tradingAgents%j.out #SBATCH --output=runs/%j/run_%j.out
#SBATCH --error=tradingAgents%j.err #SBATCH --error=runs/%j/run_%j.err
#SBATCH --time=01:00:00 #SBATCH --time=02:00:00
#SBATCH --partition=gpu #SBATCH --partition=gpu
#SBATCH --ntasks=1 #SBATCH --ntasks=1
#SBATCH --cpus-per-task=4 #SBATCH --cpus-per-task=4
#SBATCH --gpus=h100-96:2 #SBATCH --gpus=h100-96:1
#SBATCH --mem=96G #SBATCH --mem=700G
# Run the script # Run the script
uv sync uv sync

8
tradingagents/agents/utils/memory.py
View File

@ -23,18 +23,18 @@ class FinancialSituationMemory:
# instantiate the LlamaCpp embeddings wrapper # instantiate the LlamaCpp embeddings wrapper
self.embeddings_model = LlamaCppEmbeddings( self.embeddings_model = LlamaCppEmbeddings(
model_path=self.embedding, model_path=self.embedding,
n_ctx=2048, n_ctx=512,
n_parts=-1, n_parts=-1,
seed=0, seed=0,
f16_kv=True, f16_kv=True,
logits_all=False, logits_all=False,
vocab_only=False, vocab_only=False,
use_mlock=False, use_mlock=False,
n_threads=4, n_threads=16,
n_batch=512, n_batch=512,
n_gpu_layers=0, n_gpu_layers=10,
verbose=False, verbose=False,
device="cpu", device="cuda",
) )
self.client = None self.client = None
else: else:

6
tradingagents/graph/trading_graph.py
View File

@ -109,9 +109,9 @@ class TradingAgentsGraph:
elif self.config["llm_provider"].lower() == "llamacpp": elif self.config["llm_provider"].lower() == "llamacpp":
self.deep_thinking_llm = ChatLlamaCpp( self.deep_thinking_llm = ChatLlamaCpp(
model_path=self.config["deep_think_llm"], model_path=self.config["deep_think_llm"],
n_ctx=8192, n_ctx=65536,
n_batch=512, n_batch=1024,
n_gpu_layers=20, n_gpu_layers=80,
n_threads=multiprocessing.cpu_count() - 1, n_threads=multiprocessing.cpu_count() - 1,
verbose=True, verbose=True,
) )