You need to agree to share your Hugging Face information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

DeepSeek-LLM 7B telecom LoRA (legacy)

LoRA on deepseek-ai/deepseek-llm-7b-base for telecom support, network, and regulatory prose.

Legacy, incomplete eval. The private training set is not on the Hub. Trainer logs show validation loss near 0.071 through epoch 1, then collapse: train loss 0.0 and validation nan from step 4000. Treat the published adapter as an interrupted run, not a measured telecom expert.

Process-control CX: Immortalis/cx-comply.

Training (as logged)

LoRA rank 16, alpha 32, target q/k/v/o/gate/up/down
~15,756 train / 1,753 val examples claimed (unaudited)
3 planned epochs, seq 512, fp16, MPS, ~68 hours
Best finite eval loss in the table: 0.0707 at step 1800
Do not cite the later nan rows as a result

License

Inherits DeepSeek-LLM terms (license: other). The repo is gated because of that base license.

Limits

No held-out task metric (accuracy, factuality, or compliance)
Not a substitute for carrier policy engines
Prefer later CX process labels over this domain SFT

Downloads last month: -

Safetensors

Model size

7B params

Tensor type

F16

Model tree for chendren/deepseek-llm-7b-telecom-finetuned

Base model

deepseek-ai/deepseek-llm-7b-base

Adapter

(30)

this model