qwen2.5-7b-base-retool-slime-sft API & Inference Endpoint

Training

Framework: slime v0.3.0 (Megatron-LM backend).
Data: JoeYing/ReTool-SFT (2,000 multi-turn {messages} trajectories).
Recipe: loss_type=sft_loss, 3 epochs, rollout/global batch 128, Adam, lr 1e-5 (cosine → 1e-6, 10% warmup), weight decay 0.1, bf16, max_tokens_per_gpu=9216, full activation recompute.
Parallelism: 8×A100-40GB, tensor-parallel 4 (+ sequence parallel), PP 1.
Final train loss: ≈ 0.37–0.40 (from ≈ 0.70 at start).

The prompt/response template is owned by the rollout (generate_with_retool); chat templates are not applied on top during RL.

Intended use

Consume directly as the --hf-checkpoint / MODEL_PATH for the GRPO / GFlowRL stages of retool-slime, or for tool-augmented math inference using the ReTool <code>/<interpreter> format.

Training

Framework: slime v0.3.0 (Megatron-LM backend).
Data: JoeYing/ReTool-SFT (2,000 multi-turn {messages} trajectories).
Recipe: loss_type=sft_loss, 3 epochs, rollout/global batch 128, Adam, lr 1e-5 (cosine → 1e-6, 10% warmup), weight decay 0.1, bf16, max_tokens_per_gpu=9216, full activation recompute.
Parallelism: 8×A100-40GB, tensor-parallel 4 (+ sequence parallel), PP 1.
Final train loss: ≈ 0.37–0.40 (from ≈ 0.70 at start).

The prompt/response template is owned by the rollout (generate_with_retool); chat templates are not applied on top during RL.

Intended use

Consume directly as the --hf-checkpoint / MODEL_PATH for the GRPO / GFlowRL stages of retool-slime, or for tool-augmented math inference using the ReTool <code>/<interpreter> format.

qwen2.5-7b-base-retool-slime-sft

README

Training

Intended use

Explore FriendliAI today

README

Training

Intended use